Artificial Intelligence in Pediatric Dentistry: Are Chatbots Aligned With AAPD Caries Risk Assessment Guideline?

Artificial Intelligence in Pediatric Dentistry: Are Chatbots Aligned With AAPD Caries Risk Assessment Guideline? | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Artificial Intelligence in Pediatric Dentistry: Are Chatbots Aligned With AAPD Caries Risk Assessment Guideline? Esra Ceren Tuğutlu, Zeynep Betül Arslan This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9070138/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 7 You are reading this latest preprint version Abstract Objectives This study evaluated the accuracy and guideline alignment of artificial intelligence (AI)–based chatbots in pediatric caries risk assessment and management by comparing their recommendations with the American Academy of Pediatric Dentistry (AAPD) caries risk assessment guideline using simulated pediatric cases. Materials and Methods A case-based comparative study was conducted using 12 simulated pediatric patient profiles representing low, moderate, and high caries risk categories constructed according to the AAPD guideline. Five AI chatbots—ChatGPT-5.2 (Plus), ChatGPT-4o (Free), Microsoft Copilot, Google Gemini 1.5, and Claude—were evaluated. A standardized prompt instructed each chatbot to assess caries risk, recommend clinical and radiographic follow-up frequency, propose preventive interventions, and outline restorative approaches based on the AAPD guideline. Each case was presented to each chatbot on three separate days, generating 1,080 responses. Outputs were evaluated using a guideline-based scoring rubric by an experienced pediatric dentist. Statistical analyses included chi-square tests for between-chatbot comparisons and Friedman tests for within-model consistency ( p < 0.05). Results No significant differences were observed among the chatbots in caries risk classification ( p = 0.059). However, significant differences were found in clinical follow-up recommendations ( p = 0.013) and preventive interventions. Claude demonstrated higher accuracy in dietary counseling and fluoride therapy ( p = 0.001; p = 0.010), while Gemini and Copilot performed better in fissure sealant recommendations ( p = 0.006). No differences were observed in restorative treatment recommendations ( p = 0.480). Conclusions AI chatbots were generally capable of identifying pediatric caries risk levels; however, inconsistencies were observed when translating risk status into guideline-based follow-up and preventive recommendations. Clinical Relevance: AI chatbots may support dental education and preliminary clinical decision-making in pediatric dentistry, but their recommendations should be interpreted cautiously and cannot replace professional clinical judgment. artificial intelligence chatbots caries risk assessment pediatric dentistry clinical decision-making AAPD guidelines INTRODUCTION Dental caries remains one of the most prevalent chronic diseases in childhood and continues to represent a major public health concern worldwide. Untreated caries in primary teeth affects approximately 514 million children. When caries develops at an early age, it may lead to long-term oral health consequences, including pain, infection, impaired nutrition, and reduced quality of life.[ 1 ] Caries risk in children is commonly defined as the probability of developing new carious lesions within a given time frame, determined through the integration of biological susceptibility, behavioral practices, and environmental influences.[ 2 , 3 ] Dietary habits, oral hygiene practices, fluoride exposure, socioeconomic status, parental oral health behaviors, and salivary characteristics have been identified as key determinants influencing caries risk in pediatric populations.[ 2 , 4 , 5 ] Accurate and early assessment of caries risk plays a critical role in the prevention and management of dental caries in children by enabling individualized preventive and therapeutic strategies.[ 6 ] Risk-based approaches allow for the early identification of high-risk individuals who may benefit from intensified preventive measures such as more frequent follow-up visits, topical fluoride applications, dietary counseling, and parental education, while minimizing unnecessary invasive interventions in low-risk individuals.[ 3 , 7 ] To standardize and support clinical decision-making, an evidence-based guideline for pediatric caries risk assessment has been developed, most notably by the American Academy of Pediatric Dentistry (AAPD). The AAPD guideline adopts a multifactorial, risk-based framework that integrates clinical findings, caries history, dietary habits, fluoride exposure, oral hygiene practices, and social determinants of health.[ 3 ] In addition to the AAPD guideline, several other evidence-based frameworks have been developed to support caries risk assessment and management in children. Caries Management by Risk Assessment (CAMBRA) provides a structured clinical protocol that categorizes patients according to risk status and links these categories to specific preventive and therapeutic recommendations.[ 8 ] Similarly, the Cariogram model offers an algorithm-based tool that integrates biological and behavioral factors to generate a visual estimation of future caries risk, aiming to enhance objectivity and reproducibility in risk prediction.[ 9 ] While these systems rely on predefined scoring or algorithmic outputs, the AAPD guideline adopts a more flexible, clinician-centered framework that emphasizes professional judgment and contextual interpretation in translating risk indicators into individualized management plans.[ 3 ] Artificial intelligence (AI)–based chatbots utilizing large language models (LLMs) and advanced computational algorithms are increasingly being applied to interpret clinical information and support clinical decision-making in healthcare settings. Tools such as ChatGPT, Microsoft Copilot, Google Gemini, Claude and other AI models have been explored for their utility in clinical decision support tasks, including diagnostic reasoning, evidence retrieval, and treatment planning, demonstrating promising performance in systematic evaluations compared to traditional information retrieval methods.[ 10 ] Despite this potential, AI-driven systems may not yet be reliable enough to replace clinician judgment independently, and concerns persist regarding their accuracy, transparency, integration into clinical workflows, and ethical implications.[ 11 ] In particular, applications in fields requiring individualized patient assessment, such as pediatric dental practice, remain an area of active investigation to determine how AI can best complement professional expertise without compromising patient safety or clinical standards. Artificial intelligence applications in various branches of dentistry—such as endodontics, orthodontics, oral surgery, and radiology—have been extensively investigated. These studies have addressed topics including implantology, dental trauma and avulsion, pulpal-periradicular diseases, and orthodontic treatment planning.[ 12 – 15 ] However, the literature remains limited regarding AI-related research in the field of pediatric dentistry,[ 16 , 17 ] and to the best of our knowledge, no study has yet incorporated caries risk assessment in children. Therefore, the aim of this study was to evaluate and compare the accuracy and guideline alignment of AI-based chatbots in pediatric caries risk assessment, using simulated pediatric cases as the reference standard. The null hypothesis posited that AI-based chatbots would not differ significantly in the accuracy of caries risk category identification or in the diagnostics, preventive and restorative interventions they recommended. MATERIALS AND METHODS Study Design, Reporting Standards and Ethical Considerations This case-based comparative evaluation study was conducted in January 2026 in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.[ 18 ] In addition, the study was reported in line with the CHART (Chatbot Assessment Reporting Tool) Statement, which provides guidance for transparent reporting of chatbot health advice (CHA) studies.[ 19 ] As no human participants or patient-identifiable data were involved, ethical approval was not required. AI-Based Chatbots Evaluated Five AI-based chatbots were evaluated across 12 simulated pediatric dental cases. The assessed chatbots were ChatGPT-5.2 (Plus subscription, OpenAI), ChatGPT-4o (free tier, OpenAI), Microsoft Copilot (Microsoft), Google Gemini 1.5 (Google), and Claude (Anthropic). All systems are proprietary (closed-source) and are continuously updated; therefore, responses may vary over time. Reference Standard and Case Construction Simulated cases were constructed according to the American Academy of Pediatric Dentistry (AAPD) Caries-Risk Assessment and Management Manual for infants, children, and adolescents, which served as the reference standard. The AAPD guideline defines caries risk levels by integrating disease indicators, risk factors, and protective factors, and provides recommendations on individualized management pathways including recall intervals, radiographic assessment frequency, preventive interventions, and restorative treatment needs.[ 3 ] In addition, the AAPD Policy on Dietary Recommendations for Infants, Children, and Adolescents was used as a secondary reference for dietary counseling.[ 20 ] Twelve patient profiles were developed to represent low, moderate, and high caries-risk categories, incorporating a range of ages as well as diverse social, behavioral, medical, and clinical characteristics, as summarized in Table 1 . Table 1 Simulated patients representing different caries risk levels CASES AGED 0–5 YEARS Case 1 – Low Risk A 3-year-old male patient presented to the clinic for routine check-ups. The anamnesis revealed regular dental visits, tooth brushing twice daily with fluoridated toothpaste, no use of bottles containing added sugar, and no presence of active dental caries in the mother. Clinical examination showed no visible plaque or visible carious lesions. Case 2 – Moderate Risk A 5-year-old male patient presented to the clinic for routine check-ups. The anamnesis revealed regular dental visits, tooth brushing twice daily with fluoridated toothpaste, consumption of sugary snacks a maximum of twice per day, and that the family had immigrated to the country very recently. Clinical examination showed no visible plaque or white-spot lesions. Case 3 – Moderate Risk A 4-year-old female patient presented to the clinic for routine check-ups. The anamnesis revealed regular dental visits, consumption of fluoridated drinking water, and professional topical fluoride application. Extraoral examination showed a physical disability limiting the use of the child’s right arm, while intraoral examination revealed no visible plaque, no recently placed restorations, and no tooth loss due to caries. Case 4 – High Risk A 3-year-old male patient presented to the clinic for routine check-ups. The anamnesis revealed regular dental visits and tooth brushing twice daily with xylitol-containing toothpaste. In addition, the mother had active dental caries and was undergoing dental treatment, and the child has > 3 times/day exposure between-meal sugar-containing snacks. Clinical examination revealed no visible plaque or carious lesions. Case 5 – High Risk A 5-year-old female patient presented to the clinic for the fabrication of a space maintainer. The anamnesis revealed irregular dental visits, irregular tooth brushing habits, and lack of parental knowledge regarding fluoride intake. Clinical examination showed no active carious lesions or teeth lost due to caries; however, visible plaque and white-spot lesions were detected on the anterior teeth and number 75 was lost because of decay. Case 6 – High Risk A 4-year-old male patient presented to the clinic with a complaint of pain in the lower left region. The anamnesis revealed that the child consumed sugary snacks in limited amounts and only together with main meals, brushed his teeth twice daily with fluoridated toothpaste, and had irregular dental visits. Clinical examination revealed the presence of dental enamel defects and active carious lesions on teeth numbered 75, 85, as well as recently placed restorations on teeth 54, 55. CASES AGED 6 YEARS AND OLDER Case 7 – Low Risk A 7-year-old female patient presented to the clinic for routine check-ups. The anamnesis revealed tooth brushing twice daily with fluoridated toothpaste, regular dental check-ups, and professional topical fluoride application, consumption of sugary snacks a maximum of twice per day. Clinical examination showed no visible plaque. Clinical and radiographic examinations revealed no carious lesions. Case 8 – Moderate Risk A 15-year-old female patient presented to the clinic for routine check-ups. The anamnesis revealed tooth brushing twice daily with fluoridated toothpaste, regular dental visits, and the use of antidepressant medication. Clinical examination showed no visible plaque, no enamel defects, and no restorations placed within the last year. Case 9 – Moderate Risk A 9-year-old male patient presented to the clinic for routine check-ups. The patient was observed to be using a rapid maxillary expansion appliance for the treatment of maxillary constriction. The anamnesis revealed that sugar consumption was limited to once daily and taken together with main meals, and that the patient brushed his teeth twice daily with fluoridated toothpaste. Clinical and radiographic examinations revealed no carious lesions. Case 10 – High Risk A 12-year-old male patient presented to the clinic with a complaint of pain in the upper right region. The anamnesis revealed very frequent consumption of sugary beverages, irregular dental visits, and regular tooth brushing. Clinical examination revealed the patient has defective restorations on teeth 26 and 36. Bitewing radiographs taken due to suspicion of interproximal caries revealed dentin-level carious lesions on teeth 15, 24 and 36. Case 11 – High Risk An 11-year-old female patient presented to the clinic with a complaint of upper left extraoral swelling. The anamnesis revealed that the family had a low socioeconomic status, irregular dental visits, and irregular tooth brushing habits. Clinical examination revealed visible plaque, active carious lesions on teeth 16 and 26, and composite restorations placed within the last year. Case 12 – High Risk An 8-year-old female patient presented to the clinic for routine check-ups. The anamnesis revealed that the patient has irregular dental visits, irregular tooth brushing habits and frequent exposure (> 3 times/day) between-meal sugar-containing snacks or beverages per day. Clinical examination revealed no visible plaque or cavitated/non-cavitated carious lesions; however, enamel defects were present on teeth numbered 11, 21, and 26. Prompt Standardization and Query Procedure Prior to the case-based evaluation, a uniform standardized prompt was applied across all chatbot platforms to ensure comparability of responses. The chatbots were instructed to evaluate each case according to the AAPD guideline across five domains: (1) caries risk classification, (2) recommended frequency of clinical and radiographic follow-up, (3) preventive interventions (including fluoride therapy, dietary counseling, and fissure sealants), and (4) restorative interventions. The following standardized command prompt was provided to the chatbots: ‘As a pediatric dentist, evaluate the presented cases in accordance with the American Academy of Pediatric Dentistry (AAPD) guidelines by addressing the following four components: (1) assessment of the patient’s caries risk, (2) recommended frequency of clinical and radiographic follow-up examinations, (3) indicated preventive interventions, including fluoride therapy, dietary counseling, and fissure sealants, and (4) appropriate restorative treatment approaches. Evaluations should be based on the following references: American Academy of Pediatric Dentistry. Caries-risk assessment and management for infants, children, and adolescents. The Reference Manual of Pediatric Dentistry. Chicago, IL: American Academy of Pediatric Dentistry; 2025:325–331 (Tables 3 and 4 ). Dietary counseling recommendations should also follow the ‘American Academy of Pediatric Dentistry. Policy on dietary recommendations for infants, children, and adolescents. The Reference Manual of Pediatric Dentistry. Chicago, IL: American Academy of Pediatric Dentistry; 2025:118–122.’ [ 20 ] To minimize carryover effects and maintain independence between sessions, the browser history was systematically cleared after each response. Each case was presented to each chatbot on three separate days to evaluate response consistency over repeated trials. Evaluation Procedure and Reviewers All case prompts were entered into the AI-based chatbots by a researcher with expertise in oral and maxillofacial radiology. The chatbot outputs were subsequently evaluated independently by a pediatric dentist with 10 years of clinical experience, who was blinded to chatbot identity, using predefined guideline-derived criteria. In cases where the evaluator was uncertain regarding response appropriateness, the output was re-assessed by two investigators, and a final decision was reached through consensus discussion. Guideline-Based Evaluation and Scoring A scoring rubric was developed in accordance with the American Academy of Pediatric Dentistry (AAPD) guideline recommendations, including the Caries Risk Assessment and Management Manual for Infants, Children, and Adolescents (Tables 3 and 4 ) and the Policy on Dietary Recommendations for Infants, Children, and Adolescents.[ 3 , 20 ] Each chatbot response was evaluated across the five domains described above. For scoring purposes, responses that fully adhered to the guideline-based recommendations were coded as correct and complete (2). Responses that were incomplete or only partially compliant with the AAPD reference standard, or that contained erroneous or incomplete information, were coded as partially correct (1). Responses that were completely non-compliant or lacked relevant information were coded as incorrect (0). Each domain was scored independently for every case and repetition, and domain-level scores were subsequently used to calculate overall performance outcomes for each chatbot. The resulting experimental dataset comprised a total of 1,080 records (5 AI-based chatbots × 12 cases × 3 repetitions × 6 evaluation criteria). Statistical Analysis Data were analyzed using IBM SPSS Statistics version 26. For each evaluation criterion, the percentage of correct responses generated by the AI-based chatbots was calculated based on 12 clinical cases assessed across three repeated trials. Differences among chatbots in diagnostic and treatment-related responses were analyzed using the chi-square test or Fisher’s exact test, as appropriate. Friedman test was applied to compare within-model response consistency across the three assessment days. Statistical significance was set at p < 0.05. RESULTS The distribution of assessment outcomes across the five AI models is summarized in Tables 2 – 5 . Table 2 presents the responses provided for caries risk assessment, and no statistically significant difference was observed among the AI models in caries risk assessment outcomes ( p = 0.059; Table 2 ). The proportion of correct/complete responses was highest for Gemini (91.7%), followed by Copilot and Claude (both 83.3%), ChatGPT (66.7%), and ChatGPT Plus (63.9%). Incorrect assessments ranged from 8.3% to 25% across models. Table 2 Distribution of caries risk assessment outcomes according to AI models Assessment outcome ChatGPT-5.2 (Plus) n (%) ChatGPT-4o (Free) n (%) Copilot n (%) Gemini n (%) Claude n (%) p value Incorrect 9 (25) 6 (16.7) 4 (11.1) 3 (8.3) 5 (13.9) 0.059 Partially correct 4 (11.1) 6 (16.7) 2 (5.6) 0 (0) 1 (2.8) Correct/Complete 23 (63.9) 24(66.7) 30(83.3) 33 (91.7) 30(83.3) Total 36 (100) 36 (100) 36 (100) 36 (100) 36 (100) Data are expressed as n (%). Statistical analysis was performed using Fisher’s exact test. Clinical and Radiographic Follow-Up Assessment A significant difference was detected among AI models regarding clinical and radiographic follow-up assessments ( p = 0.013; Table 3 ). Gemini demonstrated the highest rate of correct/complete responses (55.6%), whereas ChatGPT Plus showed the lowest proportion (19.4%). Pairwise comparisons indicated that Gemini achieved a significantly higher proportion of correct/complete responses compared with ChatGPT Plus, ChatGPT, and Copilot (p = 0.006; p = 0.029; p = 0.016; Table 3 ). In addition, ChatGPT showed significantly higher correct/complete response rates than Copilot ( p = 0.041). No statistically significant differences were observed between the remaining chatbot pairs. ( p >0.05). Table 3 Distribution of clinical and radiographic follow-up assessment according to AI models Assessment outcome ChatGPT-5.2 (Plus) n (%) ChatGPT-4o (Free) n (%) Copilot n (%) Gemini n (%) Claude n (%) p value Incorrect 11 (30.6) 13 (36.1) 4 (11.1) 7 (19.4) 10 (27.8) 0.013* Partially correct 18 (50) 14 (38.9) 21 (58.3) 9 (25) 14 (38.9) Correct/Complete 7(19.4) ab 9 (25) a 11(30.6) b 20(55.6) c 12(33.3) abc Total 36 (100) 36 (100) 36 (100) 36 (100) 36 (100) Data are expressed as n (%). Statistical analysis was performed using Pearson’s Chi-square test. Different superscript letters (a, b, c) within the same row indicate statistically significant pairwise differences between AI models. Models sharing at least one common superscript letter do not differ significantly from each other. * indicates a statistically significant difference. Preventive Interventions Assessment Statistically significant differences were found among AI models for all preventive intervention categories (Table 4 ). For fluoride therapy interventions, the distribution of correct/complete responses differed significantly ( p = 0.010), with Claude showed the highest accuracy (36.1%), while other models demonstrated lower correct/complete rates (5.6–13.9%). Pairwise comparisons demonstrated that, for fluoride therapy interventions, Claude achieved a significantly higher proportion of correct/complete responses compared with ChatGPT Plus, Copilot, and Gemini ( p = 0.041; p = 0.003; p = 0.005; Table 4 ). No significant differences were observed among the remaining chatbots ( p >0.05). For dietary counseling interventions, Claude demonstrated a significantly higher proportion of correct/complete responses (66.6%) compared with all other AI models (0–13.9%) (p 0.05). A statistically significant difference was observed among the chatbots in fissure sealant assessments ( p = 0.006). In pairwise comparisons, ChatGPT differed significantly from Copilot, Gemini, and Claude ( p = 0.025; p = 0.007; p = 0.036). Gemini also showed significantly higher accuracy than ChatGPT Plus ( p = 0.031); no significant differences were observed between Claude and either Copilot or Gemini ( p >0.05; Table 4 ). Table 4 Distribution of preventive interventions assessment outcomes according to AI models Assessment outcome ChatGPT-5.2 (Plus) n (%) ChatGPT-4o (Free) n (%) Copilot n (%) Gemini n (%) Claude n (%) p value Fluoride therapy Incorrect 1 (2.8) 1 (2.8) 1 (2.8) 0 (0) 0 (0) 0.010* Partially correct 31 (86.1) 30 (83.3) 33 (91.7) 33 (91.7) 23 (63.9) Correct/Complete 4 (11.1) a 5(13.9) ab 2 (5.6) a 3 (8.3) a 13(36.1) b Dietary counseling Incorrect 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) < 0.001* Partially correct 33 (91.7) 31 (86.1) 34 (94.4) 36 (100) 12 (33.3) Correct/Complete 3 (8.3) a 5 (13.9) a 2 (5.6) a 0 (0) a 24 (66.6) b Fissure sealants Incorrect 4 (11.1) 8 (22.2) 1 (2.8) 0 (0) 1 (2.8) 0.006* Partially correct 12 (33.3) 8 (22.2) 6 (16.7) 7 (19.4) 13 (36.1) Correct/Complete 20 (55.6) ab 20 (55.6) a 29(80.6) bc 29(80.6) c 22(61.1) bc Data are expressed as n (%). Statistical analysis was performed using Fisher’s exact test. Different superscript letters (a, b, c) within the same row indicate statistically significant pairwise differences between AI models. Models sharing at least one common superscript letter are not statistically different. * indicates a statistically significant difference. Restorative Treatment Approaches No significant difference was observed among AI models in the assessment of restorative treatment approaches ( p = 0.480; Table 5 ). Correct/complete response rates ranged from 47.2% to 66.7%, with Claude demonstrated the highest proportion of correct/complete assessments. Table 5 Distribution of restorative treatment approaches assessment outcomes according to AI models Assessment outcome ChatGPT-5.2 (Plus) n (%) ChatGPT-4o (Free) n (%) Copilot n (%) Gemini n (%) Claude n (%) p value Incorrect 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0.480 Partially correct 19 (52.8) 18 (50) 16 (44.4) 18 (50) 12 (33.3) Correct/Complete 17 (47.2) 18 (50) 20 (55.6) 18 (50) 24 (66.7) Total 36 (100) 36 (100) 36 (100) 36 (100) 36 (100) Data are expressed as n (%). Statistical analysis was performed using Pearson’s Chi-Square test. Intra-model consistency analysis Table 6 shows the change in accuracy rates of AI-based chatbots over three days. Responses generated by ChatGPT-5.2 (Plus), ChatGPT-40 (Free), Google Gemini, and Claude did not show statistically significant differences between days ( p = 0.830; p = 0.446; p = 0.144; p = 0.202). In contrast, responses generated by Copilot showed a statistically significant variation between days ( p = 0.023). Table 6 Comparing the responses of AI programs on different days Chatbots Correct Responses 1st day n:72 (%) 2nd day n:72 (%) 3rd day n:72 (%) Total n:216 (%) p value ChatGPT-5.2 (Plus) 26 (36.1) 25 (34.7) 23 (31.9) 74 (34.3) 0.830 ChatGPT-4o (Free) 32 (44.4) 24 (33.3) 25 (34.7) 81 (37.5) 0.446 Copilot 27 (37.5) 33 (45.8) 34 (47.2) 94 (43.5) 0.023* Gemini 36 (50) 32 (44.4) 35 (48.6) 103 (47.7) 0.144 Claude 42 (58.3) 38 (52.8) 45 (62.5) 125 (57.8) 0.202 Data are expressed as n (%). The Friedman test was used to compare responses on different days. * indicates a statistically significant difference. DISCUSSION This study evaluated the guideline alignment and clinical appropriateness of five widely used AI-based chatbots in pediatric caries risk assessment and management using simulated cases constructed according to the AAPD caries risk assessment framework. To our knowledge, this is the first study to systematically assess AI chatbot performance in pediatric caries risk assessment across multiple clinical domains, including risk categorization, follow-up planning, preventive strategies, and restorative decision-making. Case-based evaluation has been recommended as an appropriate approach for assessing clinical decision support tools, as it allows standardized input, reproducibility, and direct comparison across systems. [ 10 , 11 ] Based on the results of this study, while the chatbots demonstrated comparable performance in basic caries risk categorization, significant differences emerged in follow-up planning and preventive intervention recommendations. These findings led to a partial rejection of the null hypothesis. One notable finding was the lack of statistically significant differences among AI models in caries risk categorization, while caries risk was identified with relatively high accuracy across models ( p = 0.059; Table 2 ). This indicates that contemporary large language models can identify overall caries risk levels based on structured clinical information. Similar findings have been reported in previous dental studies evaluating AI chatbots in trauma management, peri-implant diseases, and pulpal diagnosis, where models performed reasonably well in categorical diagnostic tasks.[ 12 – 14 ] In the present study, Gemini achieved the highest accuracy in identifying caries risk levels (91.7%), demonstrating superior performance in basic risk categorization compared to other models. Similarly, in the study conducted by Keleş and Arslan on dental trauma, Gemini gave the best result among chatbots with a 100% accuracy rate in diagnosis.[ 13 ] In the study conducted by Güven and colleagues on trauma, it was stated that Gemini provided more accurate and comprehensive answers also.[ 21 ] On the other hand, a pilot study on pediatric dentistry indicated that chatbots have a lower accuracy rate in diagnosis than dentists, and while they may be useful for education and patient information, they cannot replace clinicians in diagnostic decision-making processes.[ 22 ] Besides, correct risk categorization alone does not ensure appropriate patient management. In pediatric dentistry, caries risk assessment requires translating risk indicators into individualized, age-appropriate management strategies, a process that remains highly dependent on clinician expertise and contextual judgment.[ 3 , 6 ] In the current study, significant differences were detected among AI models regarding clinical and radiographic follow-up recommendations ( p = 0.013; Table 3 ). Specifically, Gemini demonstrated the highest alignment with AAPD guidelines, achieving a 'correct/complete' response rate of 55.6%, whereas ChatGPT Plus showed the lowest performance at 19.4%. This striking disparity suggests that the models' utility depends not merely on static data retrieval but on their capacity to interpret dynamic guideline tables that integrate a patient’s age, risk category, and social determinants. AAPD follow-up protocols prescribe specific recall intervals and radiographic frequencies based on risk levels (low, moderate, high), requiring a level of contextual reasoning more complex than simple categorical diagnosis.[ 3 ] The underperformance of otherwise advanced models like ChatGPT Plus in this domain may stem from insufficient specialization within the training data regarding preventive dentistry protocols or limitations in algorithmic weighting of clinical priorities. From a clinical perspective, inconsistent follow-up recommendations are concerning, as they could lead to either the progression of undetected early-stage lesions or unnecessary radiation exposure for pediatric patients, further reinforcing that chatbot outputs must remain subject to professional oversight. Preventive intervention recommendations showed the greatest heterogeneity among the evaluated AI models. In the domains of fluoride therapy and dietary counseling, Claude significantly outperformed all other chatbots ( p = 0.010; p < 0.001). While Gemini and Copilot showed higher accuracy in fissure sealant recommendations ( p = 0.006), they provided mostly partially correct advice for dietary and fluoride protocols. The high performance of Claude in dietary counseling (66.6% correct/complete) suggests a more robust integration of the AAPD Policy on Dietary Recommendations compared to its peers.[ 20 ] Conversely, the higher accuracy observed for Gemini and Copilot in fissure sealant recommendations (both 80.6% correct/complete) may reflect the relatively straightforward and clearly defined nature of AAPD guidance in this area, in contrast to dietary and fluoride recommendations, which are broader and require more contextual interpretation. These findings are consistent with broader reviews of healthcare chatbots, which often report that AI systems deliver generalized preventive advice and fail to tailor recommendations according to individual risk profiles.[ 23 ] To our knowledge, evidence on chatbot performance regarding fissure sealant and dietary recommendations is lacking; however, a prior study evaluating fluoride-related responses reported higher accuracy for Gemini compared with other models, differing from the results of the present study.[ 24 ] In another study conducted by Karamüftüoğlu et al. regarding fluoride use, ChatGPT-4.0 was found to be superior to the Gemini model.[ 25 ] The overall inconsistency across preventive categories underscores that AI systems often struggle to bridge the gap between risk identification and the formulation of a comprehensive, evidence-based management plan. According to the study results, no statistically significant difference was observed among the five AI models regarding restorative treatment approaches ( p = 0.480; Table 5 ). While Claude exhibited the highest "correct/complete" response rate at 66.7%, the success rates of the other models ranged between 47.2% and 55.6%. These findings align with the study conducted by Ozdemir and Yapici, which reported comparable performance across several large language models in responding to dentistry-specific restorative questions. [ 26 ] This consistency across models may be attributed to the well-established and prescriptive nature of restorative protocols for cavitated or dentin-level lesions, which leave less room for interpretation compared to preventive strategies. However, the fact that none of the models achieved 100% complete success, and that a significant portion of responses were only "partially correct" (ranging from 33.3% to 52.8%), proves that AI still requires professional supervision in complex clinical scenarios. In pediatric dentistry, the restorative decision-making process is dynamic, shaped not only by the presence of a lesion but also by the child's age, level of cooperation, and the anticipated exfoliation time of the tooth. Consequently, while AI models may serve as helpful auxiliary tools in restorative planning, they are not yet reliable enough to serve as autonomous decision-makers in place of clinical judgment. A significant finding of this study relates to the temporal consistency of the AI models' responses. While most models generally produced stable outputs across the three separate days of evaluation, Microsoft Copilot demonstrated a statistically significant level of daily variability ( p = 0.023; Table 6 ). Consistent with this finding, Barbosa et al. reported that Copilot exhibited the lowest consistency in their study as well.[ 12 ] This indicates that the model may provide divergent clinical recommendations over time, even when presented with identical cases. Such variability poses a serious challenge to the reliability and reproducibility of clinical decision-support mechanisms. The fact that most evaluated systems are closed-source and undergo continuous updates may be the primary reason for these temporal fluctuations. As emphasized in the literature, concerns regarding the transparency and consistency of AI-generated medical advice remain highly relevant.[ 27 – 29 ] In a clinical setting, recommending different risk categories or follow-up intervals for the same patient profile on different days could undermine standardization and jeopardize patient safety. Consequently, the fluctuations exhibited by models like Copilot reaffirm that AI tools should not yet function as standalone decision-makers but should instead be utilized as auxiliary tools under professional supervision. This study is significant as one of the few investigations in pediatric dentistry and, to our knowledge, the first to systematically compare five contemporary LLMs specifically for caries risk assessment. A major methodological strength is the use of 12 standardized AAPD-based cases evaluated over three separate days, resulting in a robust dataset of 1,080 records. Furthermore, adherence to STROBE and CHART guidelines ensures high transparency and rigor. By assessing multi-dimensional clinical decision-making, from follow-up to restorative planning, this research offers a comprehensive analysis directly relevant to pediatric dental practice. Several limitations should be considered when interpreting these findings. While the use of standardized simulated cases ensured methodological control and reproducibility, it may not fully capture the complexity of real-world clinical practice, such as behavioral variability and parental factors. Additionally, since the chatbots were explicitly instructed to follow the AAPD guidelines, their performance under alternative frameworks like CAMBRA or Cariogram remained outside the scope of this study rather than being a functional deficiency. Finally, as these are closed-source systems subject to continuous updates, the observed temporal variability indicates that the clinical reliability of these tools may fluctuate over time, reflecting a challenge inherent to current generative AI technology. CONCLUSION In conclusion, while AI chatbots demonstrate promising capability in basic pediatric caries risk categorization, significant limitations remain in translating risk status into guideline-adherent follow-up schedules and preventive strategies. Although these systems may support education and preliminary clinical decision-making, they currently lack the consistency and clinical nuance required for autonomous use in pediatric caries risk assessment and management. Until further refinement and real-world clinical validation are achieved, AI chatbots should be regarded as complementary tools rather than substitutes for professional judgment in pediatric dentistry. Declarations Ethics approval and consent to participate : Not applicable. Consent for publication: Not applicable. Data availability: The datasets analyzed during the current study are available from the corresponding author upon reasonable request. Competing interests: The authors declare no competing interests. Funding: No funding. Authors’ contributions: E.C.T. conceived the ideas; E.C.T. and Z.B.A. collected the data; Z.B.A. analysed the data; E.C.T. and Z.B.A. led the writing of the manuscript. All authors read and approved the final manuscript. Acknowledgements: None. References World Health Organization. Global oral health status report: towards universal health coverage for oral health by 2030. World Health Organization; 2022. Fejerskov O, Nyvad B, Kidd E. Dental caries: the disease and its clinical management. Wiley; 2015. American Academy of Pediatric Dentistry. Caries-risk assessment and management for infants, children, and adolescents. The Reference Manual of Pediatric Dentistry. Chicago, IL: American Academy of Pediatric Dentistry 325 – 31; 2025. American Dental Association on behalf of the Dental QualityAlliance. Guidance on caries risk assessment in children. 2018. https://www.ada.org/~/media/ADA/DQA/CRA_Report.pd-f?la=en . Accessed 4 March 2026. Issrani R, Alnusayri SAF, Alderaan DAA, Alruwaili SR, Almufarrij RAS, Alkhershawy LHS. Risk Factors for Caries in Children and Adolescents: A Systematic Review. Open Dentistry J. 2025;19:1–13. Ng TC-H, Luo BW, Lam WY-H, Baysan A, Chu C-H, Yu OY. Updates on caries risk assessment—a literature review. Dentistry J. 2024;12:312. Rechmann P, Chaffee B, Rechmann B, Featherstone J. Changes in caries risk in a practice-based randomized controlled trial. Adv Dent Res. 2018;29:15–23. Featherstone JD, Alston P, Chaffee BW, Rechmann P. Caries Management by Risk Assessment (CAMBRA): an update for use in clinical practice for patients aged 6 through adult. J Calif Dent Assoc. 2019;47:25–34. Bratthall D, Hänsel Petersson G. Cariogram–a multifactorial risk assessment model for a multifactorial disease. Community Dent Oral Epidemiol. 2005;33:256–64. 10.1111/j.1600-0528.2005.00233.x . Sandmann S, Riepenhausen S, Plagwitz L, Varghese J. Systematic analysis of ChatGPT, Google search and Llama 2 for clinical decision support tasks. Nat Commun. 2024;15(1):2050. Deng J, Heybati K, Park Y-J, Zhou F, Bozzo A. Artificial intelligence in clinical practice: a look at ChatGPT. Cleve Clin J Med. 2024;91:173–80. Barbosa IA, Alves MSA, de Almeida PRZ, de Almeida Rodrigues P, de Oliveira RP, de Menezes SAF, de Moura JDM, de Souza Fonseca RR. (2025) Assessing the diagnostic and treatment accuracy of large Language models (LLMs) in Peri-Implant diseases: a clinical experimental study. J Dent:106091. Keleş ÖK, Arslan ZB. Performance of artificial intelligence chatbots in the diagnosis and management of simulated dental trauma cases: an evaluation based on IADT guidelines. Clin Oral Invest. 2026;30:26. de Moura JDM, Fontana CE, da Silva Lima VHR, de Souza Alves I, de Melo Santos PA, de Almeida Rodrigues P. Comparative accuracy of artificial intelligence chatbots in pulpal and periradicular diagnosis: A cross-sectional study. Comput Biol Med. 2024;183:109332. Mustuloğlu Ş, Deniz BP. Evaluation of chatbots in the emergency management of avulsion injuries. Dent Traumatol. 2025;41(4):437–44. Bayraktar Nahir C. Can ChatGPT be guide in pediatric dentistry? BMC Oral Health. 2025;25:9. 10.1186/s12903-024-05393-1 . Kaya İ, Demirel A. Evaluation of Accuracy, Information Quality, and Readability of Artificial Intelligence Based Chatbots in Pediatric Oral Surgery: A Comparative Analysis Based on the AAPD Clinical Guideline. Cumhuriyet Dent J. 2025;28:586–93. Malta M, Cardoso LO, Bastos FI, Magnanini MMF, Silva, CMFPd. STROBE initiative: guidelines on reporting observational studies. Rev Saude Publica. 2010;44:559–65. Huo B, Collins GS, Chartash D, Thirunavukarasu AJ, Flanagin A, Iorio A, Cacciamani G, Chen X, Liu N, Mathur P. Reporting guideline for chatbot health advice studies: the CHART statement. JAMA Netw open. 2025;8:e2530220–2530220. American Academy of Pediatric Dentistry. Policy on dietary recommendations for infants, children, and adolescents. The Reference Manual of Pediatric Dentistry. Chicago, IL: American Academy of Pediatric Dentistry 118 – 22; 2025. Guven Y, Ozdemir OT, Kavan MY. Performance of Artificial Intelligence Chatbots in Responding to Patient Queries Related to Traumatic Dental Injuries: A Comparative Study. Dent Traumatol. 2025;41:338–47. 10.1111/edt.13020 . Rokhshad R, Zhang P, Mohammad-Rahimi H, Pitchika V, Entezari N, Schwendicke F. Accuracy and consistency of chatbots versus clinicians for answering pediatric dentistry questions: A pilot study. J Dent. 2024;144:104938. Biswas S, Biswas S. (2023) Title: Role of ChatGPT in Dental Science. Available at SSRN 4403581, 2023. Accessed 4 March 2026. Aksu S, Bakır FA. Yapay zeka sohbet robotlarının ‘diş hekimliğinde flor’konusu ile ilgili sorulara verdikleri yanıtların değerlendirilmesi. Mersin Üniversitesi Tıp Fakültesi Lokman Hekim Tıp. Tarihi ve Folklorik Tıp Dergisi. 2025;15:1069–78. Karamüftüoğlu N, Varol EA, Bal C. Exploring artificial intelligence chatbots in pediatric fluoride education: a cross-sectional study. Sci Rep. 2025;16:182. 10.1038/s41598-025-28857-y . Ozdemir ZM, Yapici E. Evaluating the accuracy, reliability, consistency, and readability of different large language models in restorative dentistry. J Esthetic Restor Dentistry. 2025;37:1740–52. Makrygiannakis MA, Giannakopoulos K, Kaklamanos EG. Evidence-based potential of generative artificial intelligence large language models in orthodontics: a comparative study of ChatGPT, Google Bard, and Microsoft Bing. Eur J Orthod. 2026;48:cjae017. Dermata A, Arhakis A, Makrygiannakis MA, Giannakopoulos K, Kaklamanos EG. Evaluating the evidence-based potential of six large language models in paediatric dentistry: a comparative study on generative artificial intelligence. Eur Archives Pediatr Dentistry. 2025;26:527–35. Alhaidry HM, Fatani B, Alrayes JO, Almana AM, Alfhaed NK, Alhaidry H, Alrayes J, Almana A and Alfhaed, Sr NK. (2023) ChatGPT in dentistry: a comprehensive review. Cureus 15(4): e38317. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Reviews received at journal 20 Apr, 2026 Reviewers agreed at journal 08 Apr, 2026 Reviewers invited by journal 17 Mar, 2026 Editor invited by journal 12 Mar, 2026 Editor assigned by journal 11 Mar, 2026 Submission checks completed at journal 11 Mar, 2026 First submitted to journal 09 Mar, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9070138","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":607654371,"identity":"1d4ec875-c30c-4e7a-a6e9-89a04fcff67f","order_by":0,"name":"Esra Ceren Tuğutlu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA1ElEQVRIiWNgGAWjYNCCAgk5BgYeEIuZGOVARQcMJIwRWtiI08KQ2EC0Fn728wc/fzCwSO+fkXtMgqHCOrFBvvcBXi2SPcnMEkCH5c64kZcmwXAmPbGBjd0ArxaDG8wMYC0NN3LMJBjbDgO1EHCZ/Q1m5h9ALenyYC3/iNBiIMHMBrIlwQCspYEILRJnks0szhhIGG488y7ZIuFYunEbWxp+LfztBx/fqKiok5c7nnvwxocaa9l+5mP4tSCAQAIDAxAREZMI+w4Qr3YUjIJRMApGFgAAjdA+p+A69AsAAAAASUVORK5CYII=","orcid":"","institution":"Ankara Yıldırım Beyazıt University","correspondingAuthor":true,"prefix":"","firstName":"Esra","middleName":"Ceren","lastName":"Tuğutlu","suffix":""},{"id":607654372,"identity":"f1ec7a33-de0b-4708-814d-3d584d6596a8","order_by":1,"name":"Zeynep Betül Arslan","email":"","orcid":"","institution":"Ankara Yıldırım Beyazıt University","correspondingAuthor":false,"prefix":"","firstName":"Zeynep","middleName":"Betül","lastName":"Arslan","suffix":""}],"badges":[],"createdAt":"2026-03-09 08:09:22","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9070138/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9070138/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":105035493,"identity":"b1872a63-abef-42ca-bae9-b3315c860d2c","added_by":"auto","created_at":"2026-03-20 07:26:12","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1146845,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9070138/v1/7d0ba724-341c-4d6c-b0ac-55b5642b407f.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"\u003cp\u003eArtificial Intelligence in Pediatric Dentistry: Are Chatbots Aligned With AAPD Caries Risk Assessment Guideline?\u003c/p\u003e","fulltext":[{"header":"INTRODUCTION","content":"\u003cp\u003eDental caries remains one of the most prevalent chronic diseases in childhood and continues to represent a major public health concern worldwide. Untreated caries in primary teeth affects approximately 514\u0026nbsp;million children. When caries develops at an early age, it may lead to long-term oral health consequences, including pain, infection, impaired nutrition, and reduced quality of life.[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]\u003c/p\u003e \u003cp\u003eCaries risk in children is commonly defined as the probability of developing new carious lesions within a given time frame, determined through the integration of biological susceptibility, behavioral practices, and environmental influences.[\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e] Dietary habits, oral hygiene practices, fluoride exposure, socioeconomic status, parental oral health behaviors, and salivary characteristics have been identified as key determinants influencing caries risk in pediatric populations.[\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e] Accurate and early assessment of caries risk plays a critical role in the prevention and management of dental caries in children by enabling individualized preventive and therapeutic strategies.[\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e] Risk-based approaches allow for the early identification of high-risk individuals who may benefit from intensified preventive measures such as more frequent follow-up visits, topical fluoride applications, dietary counseling, and parental education, while minimizing unnecessary invasive interventions in low-risk individuals.[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]\u003c/p\u003e \u003cp\u003e To standardize and support clinical decision-making, an evidence-based guideline for pediatric caries risk assessment has been developed, most notably by the American Academy of Pediatric Dentistry (AAPD). The AAPD guideline adopts a multifactorial, risk-based framework that integrates clinical findings, caries history, dietary habits, fluoride exposure, oral hygiene practices, and social determinants of health.[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e] In addition to the AAPD guideline, several other evidence-based frameworks have been developed to support caries risk assessment and management in children. Caries Management by Risk Assessment (CAMBRA) provides a structured clinical protocol that categorizes patients according to risk status and links these categories to specific preventive and therapeutic recommendations.[\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e] Similarly, the Cariogram model offers an algorithm-based tool that integrates biological and behavioral factors to generate a visual estimation of future caries risk, aiming to enhance objectivity and reproducibility in risk prediction.[\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e] While these systems rely on predefined scoring or algorithmic outputs, the AAPD guideline adopts a more flexible, clinician-centered framework that emphasizes professional judgment and contextual interpretation in translating risk indicators into individualized management plans.[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]\u003c/p\u003e \u003cp\u003eArtificial intelligence (AI)\u0026ndash;based chatbots utilizing large language models (LLMs) and advanced computational algorithms are increasingly being applied to interpret clinical information and support clinical decision-making in healthcare settings. Tools such as ChatGPT, Microsoft Copilot, Google Gemini, Claude and other AI models have been explored for their utility in clinical decision support tasks, including diagnostic reasoning, evidence retrieval, and treatment planning, demonstrating promising performance in systematic evaluations compared to traditional information retrieval methods.[\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e] Despite this potential, AI-driven systems may not yet be reliable enough to replace clinician judgment independently, and concerns persist regarding their accuracy, transparency, integration into clinical workflows, and ethical implications.[\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e] In particular, applications in fields requiring individualized patient assessment, such as pediatric dental practice, remain an area of active investigation to determine how AI can best complement professional expertise without compromising patient safety or clinical standards.\u003c/p\u003e \u003cp\u003eArtificial intelligence applications in various branches of dentistry\u0026mdash;such as endodontics, orthodontics, oral surgery, and radiology\u0026mdash;have been extensively investigated. These studies have addressed topics including implantology, dental trauma and avulsion, pulpal-periradicular diseases, and orthodontic treatment planning.[\u003cspan additionalcitationids=\"CR13 CR14\" citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e] However, the literature remains limited regarding AI-related research in the field of pediatric dentistry,[\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e] and to the best of our knowledge, no study has yet incorporated caries risk assessment in children. Therefore, the aim of this study was to evaluate and compare the accuracy and guideline alignment of AI-based chatbots in pediatric caries risk assessment, using simulated pediatric cases as the reference standard. The null hypothesis posited that AI-based chatbots would not differ significantly in the accuracy of caries risk category identification or in the diagnostics, preventive and restorative interventions they recommended.\u003c/p\u003e"},{"header":"MATERIALS AND METHODS","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eStudy Design, Reporting Standards and Ethical Considerations\u003c/h2\u003e \u003cp\u003e This case-based comparative evaluation study was conducted in January 2026 in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.[\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e] In addition, the study was reported in line with the CHART (Chatbot Assessment Reporting Tool) Statement, which provides guidance for transparent reporting of chatbot health advice (CHA) studies.[\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e] As no human participants or patient-identifiable data were involved, ethical approval was not required.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003e\u003c/h3\u003e\n\u003cdiv class=\"Heading\"\u003e\u003cb\u003eAI-Based Chatbots Evaluated\u003c/b\u003e\u003c/div\u003e \u003cp\u003eFive AI-based chatbots were evaluated across 12 simulated pediatric dental cases. The assessed chatbots were ChatGPT-5.2 (Plus subscription, OpenAI), ChatGPT-4o (free tier, OpenAI), Microsoft Copilot (Microsoft), Google Gemini 1.5 (Google), and Claude (Anthropic). All systems are proprietary (closed-source) and are continuously updated; therefore, responses may vary over time.\u003c/p\u003e \u003cp\u003e \u003cb\u003eReference Standard and Case Construction\u003c/b\u003e \u003c/p\u003e \u003cp\u003e Simulated cases were constructed according to the American Academy of Pediatric Dentistry (AAPD) Caries-Risk Assessment and Management Manual for infants, children, and adolescents, which served as the reference standard. The AAPD guideline defines caries risk levels by integrating disease indicators, risk factors, and protective factors, and provides recommendations on individualized management pathways including recall intervals, radiographic assessment frequency, preventive interventions, and restorative treatment needs.[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e] In addition, the AAPD Policy on Dietary Recommendations for Infants, Children, and Adolescents was used as a secondary reference for dietary counseling.[\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]\u003c/p\u003e \u003cp\u003eTwelve patient profiles were developed to represent low, moderate, and high caries-risk categories, incorporating a range of ages as well as diverse social, behavioral, medical, and clinical characteristics, as summarized in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eSimulated patients representing different caries risk levels\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"2\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eCASES AGED 0\u0026ndash;5 YEARS\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eCase 1 \u0026ndash; Low Risk\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eA 3-year-old male patient presented to the clinic for routine check-ups. The anamnesis revealed regular dental visits, tooth brushing twice daily with fluoridated toothpaste, no use of bottles containing added sugar, and no presence of active dental caries in the mother. Clinical examination showed no visible plaque or visible carious lesions.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eCase 2 \u0026ndash; Moderate Risk\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eA 5-year-old male patient presented to the clinic for routine check-ups. The anamnesis revealed regular dental visits, tooth brushing twice daily with fluoridated toothpaste, consumption of sugary snacks a maximum of twice per day, and that the family had immigrated to the country very recently. Clinical examination showed no visible plaque or white-spot lesions.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eCase 3 \u0026ndash; Moderate Risk\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eA 4-year-old female patient presented to the clinic for routine check-ups. The anamnesis revealed regular dental visits, consumption of fluoridated drinking water, and professional topical fluoride application. Extraoral examination showed a physical disability limiting the use of the child\u0026rsquo;s right arm, while intraoral examination revealed no visible plaque, no recently placed restorations, and no tooth loss due to caries.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eCase 4 \u0026ndash; High Risk\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eA 3-year-old male patient presented to the clinic for routine check-ups. The anamnesis revealed regular dental visits and tooth brushing twice daily with xylitol-containing toothpaste. In addition, the mother had active dental caries and was undergoing dental treatment, and the child has \u0026gt;\u0026thinsp;3 times/day exposure between-meal sugar-containing snacks. Clinical examination revealed no visible plaque or carious lesions.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eCase 5 \u0026ndash; High Risk\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eA 5-year-old female patient presented to the clinic for the fabrication of a space maintainer. The anamnesis revealed irregular dental visits, irregular tooth brushing habits, and lack of parental knowledge regarding fluoride intake. Clinical examination showed no active carious lesions or teeth lost due to caries; however, visible plaque and white-spot lesions were detected on the anterior teeth and number 75 was lost because of decay.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eCase 6 \u0026ndash; High Risk\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eA 4-year-old male patient presented to the clinic with a complaint of pain in the lower left region. The anamnesis revealed that the child consumed sugary snacks in limited amounts and only together with main meals, brushed his teeth twice daily with fluoridated toothpaste, and had irregular dental visits. Clinical examination revealed the presence of dental enamel defects and active carious lesions on teeth numbered 75, 85, as well as recently placed restorations on teeth 54, 55.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eCASES AGED 6 YEARS AND OLDER\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eCase 7 \u0026ndash; Low Risk\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eA 7-year-old female patient presented to the clinic for routine check-ups. The anamnesis revealed tooth brushing twice daily with fluoridated toothpaste, regular dental check-ups, and professional topical fluoride application, consumption of sugary snacks a maximum of twice per day. Clinical examination showed no visible plaque. Clinical and radiographic examinations revealed no carious lesions.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eCase 8 \u0026ndash; Moderate Risk\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eA 15-year-old female patient presented to the clinic for routine check-ups. The anamnesis revealed tooth brushing twice daily with fluoridated toothpaste, regular dental visits, and the use of antidepressant medication. Clinical examination showed no visible plaque, no enamel defects, and no restorations placed within the last year.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eCase 9 \u0026ndash; Moderate Risk\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eA 9-year-old male patient presented to the clinic for routine check-ups. The patient was observed to be using a rapid maxillary expansion appliance for the treatment of maxillary constriction. The anamnesis revealed that sugar consumption was limited to once daily and taken together with main meals, and that the patient brushed his teeth twice daily with fluoridated toothpaste. Clinical and radiographic examinations revealed no carious lesions.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eCase 10 \u0026ndash; High Risk\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eA 12-year-old male patient presented to the clinic with a complaint of pain in the upper right region. The anamnesis revealed very frequent consumption of sugary beverages, irregular dental visits, and regular tooth brushing. Clinical examination revealed the patient has defective restorations on teeth 26 and 36. Bitewing radiographs taken due to suspicion of interproximal caries revealed dentin-level carious lesions on teeth 15, 24 and 36.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eCase 11 \u0026ndash; High Risk\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAn 11-year-old female patient presented to the clinic with a complaint of upper left extraoral swelling. The anamnesis revealed that the family had a low socioeconomic status, irregular dental visits, and irregular tooth brushing habits. Clinical examination revealed visible plaque, active carious lesions on teeth 16 and 26, and composite restorations placed within the last year.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eCase 12 \u0026ndash; High Risk\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAn 8-year-old female patient presented to the clinic for routine check-ups. The anamnesis revealed that the patient has irregular dental visits, irregular tooth brushing habits and frequent exposure (\u0026gt;\u0026thinsp;3 times/day) between-meal sugar-containing snacks or beverages per day. Clinical examination revealed no visible plaque or cavitated/non-cavitated carious lesions; however, enamel defects were present on teeth numbered 11, 21, and 26.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e\n\u003ch3\u003ePrompt Standardization and Query Procedure\u003c/h3\u003e\n\u003cp\u003ePrior to the case-based evaluation, a uniform standardized prompt was applied across all chatbot platforms to ensure comparability of responses. The chatbots were instructed to evaluate each case according to the AAPD guideline across five domains: (1) caries risk classification, (2) recommended frequency of clinical and radiographic follow-up, (3) preventive interventions (including fluoride therapy, dietary counseling, and fissure sealants), and (4) restorative interventions.\u003c/p\u003e \u003cp\u003e The following standardized command prompt was provided to the chatbots: \u0026lsquo;As a pediatric dentist, evaluate the presented cases in accordance with the American Academy of Pediatric Dentistry (AAPD) guidelines by addressing the following four components: (1) assessment of the patient\u0026rsquo;s caries risk, (2) recommended frequency of clinical and radiographic follow-up examinations, (3) indicated preventive interventions, including fluoride therapy, dietary counseling, and fissure sealants, and (4) appropriate restorative treatment approaches. Evaluations should be based on the following references: American Academy of Pediatric Dentistry. Caries-risk assessment and management for infants, children, and adolescents. The Reference Manual of Pediatric Dentistry. Chicago, IL: American Academy of Pediatric Dentistry; 2025:325\u0026ndash;331 (Tables\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e and \u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). Dietary counseling recommendations should also follow the \u0026lsquo;American Academy of Pediatric Dentistry. Policy on dietary recommendations for infants, children, and adolescents. The Reference Manual of Pediatric Dentistry. Chicago, IL: American Academy of Pediatric Dentistry; 2025:118\u0026ndash;122.\u0026rsquo; [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]\u003c/p\u003e \u003cp\u003eTo minimize carryover effects and maintain independence between sessions, the browser history was systematically cleared after each response. Each case was presented to each chatbot on three separate days to evaluate response consistency over repeated trials.\u003c/p\u003e\n\u003ch3\u003eEvaluation Procedure and Reviewers\u003c/h3\u003e\n\u003cp\u003eAll case prompts were entered into the AI-based chatbots by a researcher with expertise in oral and maxillofacial radiology. The chatbot outputs were subsequently evaluated independently by a pediatric dentist with 10 years of clinical experience, who was blinded to chatbot identity, using predefined guideline-derived criteria. In cases where the evaluator was uncertain regarding response appropriateness, the output was re-assessed by two investigators, and a final decision was reached through consensus discussion.\u003c/p\u003e\n\u003ch3\u003eGuideline-Based Evaluation and Scoring\u003c/h3\u003e\n\u003cp\u003eA scoring rubric was developed in accordance with the American Academy of Pediatric Dentistry (AAPD) guideline recommendations, including the Caries Risk Assessment and Management Manual for Infants, Children, and Adolescents (Tables\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e and \u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e) and the Policy on Dietary Recommendations for Infants, Children, and Adolescents.[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]\u003c/p\u003e \u003cp\u003eEach chatbot response was evaluated across the five domains described above. For scoring purposes, responses that fully adhered to the guideline-based recommendations were coded as correct and complete (2). Responses that were incomplete or only partially compliant with the AAPD reference standard, or that contained erroneous or incomplete information, were coded as partially correct (1). Responses that were completely non-compliant or lacked relevant information were coded as incorrect (0).\u003c/p\u003e \u003cp\u003eEach domain was scored independently for every case and repetition, and domain-level scores were subsequently used to calculate overall performance outcomes for each chatbot. The resulting experimental dataset comprised a total of 1,080 records (5 AI-based chatbots \u0026times; 12 cases \u0026times; 3 repetitions \u0026times; 6 evaluation criteria).\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eStatistical Analysis\u003c/h2\u003e \u003cp\u003eData were analyzed using IBM SPSS Statistics version 26. For each evaluation criterion, the percentage of correct responses generated by the AI-based chatbots was calculated based on 12 clinical cases assessed across three repeated trials. Differences among chatbots in diagnostic and treatment-related responses were analyzed using the chi-square test or Fisher\u0026rsquo;s exact test, as appropriate. Friedman test was applied to compare within-model response consistency across the three assessment days. Statistical significance was set at p\u0026thinsp;\u0026lt;\u0026thinsp;0.05.\u003c/p\u003e \u003c/div\u003e"},{"header":"RESULTS","content":"\u003cp\u003eThe distribution of assessment outcomes across the five AI models is summarized in Tables\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e\u0026ndash;\u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e5\u003c/span\u003e. Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e presents the responses provided for caries risk assessment, and no statistically significant difference was observed among the AI models in caries risk assessment outcomes (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.059; Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). The proportion of correct/complete responses was highest for Gemini (91.7%), followed by Copilot and Claude (both 83.3%), ChatGPT (66.7%), and ChatGPT Plus (63.9%). Incorrect assessments ranged from 8.3% to 25% across models.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eDistribution of caries risk assessment outcomes according to AI models\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"7\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAssessment outcome\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eChatGPT-5.2 (Plus)\u003c/p\u003e \u003cp\u003en (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eChatGPT-4o (Free) n (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eCopilot\u003c/p\u003e \u003cp\u003en (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eGemini\u003c/p\u003e \u003cp\u003en (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eClaude\u003c/p\u003e \u003cp\u003en (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003e\u003cem\u003ep\u003c/em\u003e value\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIncorrect\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e9 (25)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e6 (16.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e4 (11.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e3 (8.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e5 (13.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\" morerows=\"3\" rowspan=\"4\"\u003e \u003cp\u003e0.059\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePartially correct\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4 (11.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e6 (16.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2 (5.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0 (0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1 (2.8)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCorrect/Complete\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e23 (63.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e24(66.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e30(83.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e33 (91.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e30(83.3)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTotal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e36 (100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e36 (100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e36 (100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e36 (100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e36 (100)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eData are expressed as n (%). Statistical analysis was performed using Fisher\u0026rsquo;s exact test.\u003c/p\u003e\n\u003ch3\u003eClinical and Radiographic Follow-Up Assessment\u003c/h3\u003e\n\u003cp\u003eA significant difference was detected among AI models regarding clinical and radiographic follow-up assessments (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.013; Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). Gemini demonstrated the highest rate of correct/complete responses (55.6%), whereas ChatGPT Plus showed the lowest proportion (19.4%).\u003c/p\u003e \u003cp\u003ePairwise comparisons indicated that Gemini achieved a significantly higher proportion of correct/complete responses compared with ChatGPT Plus, ChatGPT, and Copilot (p\u0026thinsp;=\u0026thinsp;0.006; \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.029; \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.016; Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). In addition, ChatGPT showed significantly higher correct/complete response rates than Copilot (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.041). No statistically significant differences were observed between the remaining chatbot pairs. (\u003cem\u003ep\u003c/em\u003e\u0026gt;0.05).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eDistribution of clinical and radiographic follow-up assessment according to AI models\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"7\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAssessment outcome\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eChatGPT-5.2 (Plus)\u003c/p\u003e \u003cp\u003en (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eChatGPT-4o (Free) n (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eCopilot\u003c/p\u003e \u003cp\u003en (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eGemini\u003c/p\u003e \u003cp\u003en (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eClaude\u003c/p\u003e \u003cp\u003en (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003e\u003cem\u003ep\u003c/em\u003e value\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIncorrect\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e11 (30.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e13 (36.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e4 (11.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e7 (19.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e10 (27.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\" morerows=\"3\" rowspan=\"4\"\u003e \u003cp\u003e0.013*\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePartially correct\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e18 (50)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e14 (38.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e21 (58.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e9 (25)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e14 (38.9)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCorrect/Complete\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e7(19.4)\u003csup\u003eab\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e9 (25)\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e11(30.6)\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e20(55.6)\u003csup\u003ec\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e12(33.3)\u003csup\u003eabc\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTotal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e36 (100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e36 (100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e36 (100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e36 (100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e36 (100)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eData are expressed as n (%). Statistical analysis was performed using Pearson\u0026rsquo;s Chi-square test. Different superscript letters (a, b, c) within the same row indicate statistically significant pairwise differences between AI models. Models sharing at least one common superscript letter do not differ significantly from each other. * indicates a statistically significant difference.\u003c/p\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003ePreventive Interventions Assessment\u003c/h2\u003e \u003cp\u003eStatistically significant differences were found among AI models for all preventive intervention categories (Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eFor fluoride therapy interventions, the distribution of correct/complete responses differed significantly (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.010), with Claude showed the highest accuracy (36.1%), while other models demonstrated lower correct/complete rates (5.6\u0026ndash;13.9%). Pairwise comparisons demonstrated that, for fluoride therapy interventions, Claude achieved a significantly higher proportion of correct/complete responses compared with ChatGPT Plus, Copilot, and Gemini (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.041; \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.003; \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.005; Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). No significant differences were observed among the remaining chatbots (\u003cem\u003ep\u003c/em\u003e\u0026gt;0.05).\u003c/p\u003e \u003cp\u003eFor dietary counseling interventions, Claude demonstrated a significantly higher proportion of correct/complete responses (66.6%) compared with all other AI models (0\u0026ndash;13.9%) (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001; Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e), while no significant differences were observed among ChatGPT Plus, ChatGPT, Copilot, and Gemini (\u003cem\u003ep\u003c/em\u003e\u0026gt;0.05).\u003c/p\u003e \u003cp\u003eA statistically significant difference was observed among the chatbots in fissure sealant assessments (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.006). In pairwise comparisons, ChatGPT differed significantly from Copilot, Gemini, and Claude (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.025; \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.007; \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.036). Gemini also showed significantly higher accuracy than ChatGPT Plus (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.031); no significant differences were observed between Claude and either Copilot or Gemini (\u003cem\u003ep\u003c/em\u003e\u0026gt;0.05; Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eDistribution of preventive interventions assessment outcomes according to AI models\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"7\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAssessment outcome\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eChatGPT-5.2 (Plus)\u003c/p\u003e \u003cp\u003en (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eChatGPT-4o (Free) n (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eCopilot\u003c/p\u003e \u003cp\u003en (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eGemini\u003c/p\u003e \u003cp\u003en (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eClaude\u003c/p\u003e \u003cp\u003en (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003e\u003cem\u003ep\u003c/em\u003e value\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colspan=\"7\" nameend=\"c7\" namest=\"c1\"\u003e \u003cp\u003eFluoride therapy\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIncorrect\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1 (2.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1 (2.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1 (2.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0 (0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0 (0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\" morerows=\"2\" rowspan=\"3\"\u003e \u003cp\u003e0.010*\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePartially correct\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e31 (86.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e30 (83.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e33 (91.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e33 (91.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e23 (63.9)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCorrect/Complete\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4 (11.1)\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e5(13.9)\u003csup\u003eab\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2 (5.6)\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e3 (8.3)\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e13(36.1)\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"7\" nameend=\"c7\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eDietary counseling\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIncorrect\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0 (0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0 (0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0 (0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0 (0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0 (0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\" morerows=\"2\" rowspan=\"3\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001*\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePartially correct\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e33 (91.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e31 (86.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e34 (94.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e36 (100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e12 (33.3)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCorrect/Complete\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3 (8.3)\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e5 (13.9)\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2 (5.6)\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0 (0)\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e24 (66.6)\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"7\" nameend=\"c7\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eFissure sealants\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIncorrect\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4 (11.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e8 (22.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1 (2.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0 (0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1 (2.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\" morerows=\"2\" rowspan=\"3\"\u003e \u003cp\u003e0.006*\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePartially correct\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e12 (33.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e8 (22.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e6 (16.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e7 (19.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e13 (36.1)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCorrect/Complete\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e20 (55.6)\u003csup\u003eab\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e20 (55.6)\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e29(80.6)\u003csup\u003ebc\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e29(80.6)\u003csup\u003ec\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e22(61.1)\u003csup\u003ebc\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eData are expressed as n (%). Statistical analysis was performed using Fisher\u0026rsquo;s exact test. Different superscript letters (a, b, c) within the same row indicate statistically significant pairwise differences between AI models. Models sharing at least one common superscript letter are not statistically different. \u003cb\u003e*\u003c/b\u003e indicates a statistically significant difference.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eRestorative Treatment Approaches\u003c/h2\u003e \u003cp\u003eNo significant difference was observed among AI models in the assessment of restorative treatment approaches (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.480; Table\u0026nbsp;\u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e5\u003c/span\u003e). Correct/complete response rates ranged from 47.2% to 66.7%, with Claude demonstrated the highest proportion of correct/complete assessments.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab5\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 5\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eDistribution of restorative treatment approaches assessment outcomes according to AI models\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"7\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAssessment outcome\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eChatGPT-5.2 (Plus)\u003c/p\u003e \u003cp\u003en (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eChatGPT-4o (Free) n (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eCopilot\u003c/p\u003e \u003cp\u003en (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eGemini\u003c/p\u003e \u003cp\u003en (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eClaude\u003c/p\u003e \u003cp\u003en (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003e\u003cem\u003ep\u003c/em\u003e value\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIncorrect\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0 (0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0 (0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0 (0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0 (0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0 (0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\" morerows=\"3\" rowspan=\"4\"\u003e \u003cp\u003e0.480\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePartially correct\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e19 (52.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e18 (50)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e16 (44.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e18 (50)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e12 (33.3)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCorrect/Complete\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e17 (47.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e18 (50)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e20 (55.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e18 (50)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e24 (66.7)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTotal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e36 (100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e36 (100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e36 (100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e36 (100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e36 (100)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eData are expressed as n (%). Statistical analysis was performed using Pearson\u0026rsquo;s Chi-Square test.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eIntra-model consistency analysis\u003c/h2\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab6\" class=\"InternalRef\"\u003e6\u003c/span\u003e shows the change in accuracy rates of AI-based chatbots over three days. Responses generated by ChatGPT-5.2 (Plus), ChatGPT-40 (Free), Google Gemini, and Claude did not show statistically significant differences between days (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.830; \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.446; \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.144; \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.202). In contrast, responses generated by Copilot showed a statistically significant variation between days (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.023).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab6\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 6\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eComparing the responses of AI programs on different days\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eChatbots\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"4\" nameend=\"c5\" namest=\"c2\"\u003e \u003cp\u003eCorrect Responses\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1st day\u003c/p\u003e \u003cp\u003en:72 (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2nd day\u003c/p\u003e \u003cp\u003en:72 (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e3rd day\u003c/p\u003e \u003cp\u003en:72 (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eTotal\u003c/p\u003e \u003cp\u003en:216 (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cem\u003ep\u003c/em\u003e value\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eChatGPT-5.2 (Plus)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e26 (36.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e25 (34.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e23 (31.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e74 (34.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.830\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eChatGPT-4o (Free)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e32 (44.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e24 (33.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e25 (34.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e81 (37.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.446\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eCopilot\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e27 (37.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e33 (45.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e34 (47.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e94 (43.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.023*\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eGemini\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e36 (50)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e32 (44.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e35 (48.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e103 (47.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.144\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eClaude\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e42 (58.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e38 (52.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e45 (62.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e125 (57.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.202\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eData are expressed as n (%). The Friedman test was used to compare responses on different days. \u003cb\u003e*\u003c/b\u003e indicates a statistically significant difference.\u003c/p\u003e \u003c/div\u003e"},{"header":"DISCUSSION","content":"\u003cp\u003e This study evaluated the guideline alignment and clinical appropriateness of five widely used AI-based chatbots in pediatric caries risk assessment and management using simulated cases constructed according to the AAPD caries risk assessment framework. To our knowledge, this is the first study to systematically assess AI chatbot performance in pediatric caries risk assessment across multiple clinical domains, including risk categorization, follow-up planning, preventive strategies, and restorative decision-making. Case-based evaluation has been recommended as an appropriate approach for assessing clinical decision support tools, as it allows standardized input, reproducibility, and direct comparison across systems. [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e] Based on the results of this study, while the chatbots demonstrated comparable performance in basic caries risk categorization, significant differences emerged in follow-up planning and preventive intervention recommendations. These findings led to a partial rejection of the null hypothesis.\u003c/p\u003e \u003cp\u003eOne notable finding was the lack of statistically significant differences among AI models in caries risk categorization, while caries risk was identified with relatively high accuracy across models (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.059; Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). This indicates that contemporary large language models can identify overall caries risk levels based on structured clinical information. Similar findings have been reported in previous dental studies evaluating AI chatbots in trauma management, peri-implant diseases, and pulpal diagnosis, where models performed reasonably well in categorical diagnostic tasks.[\u003cspan additionalcitationids=\"CR13\" citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e] In the present study, Gemini achieved the highest accuracy in identifying caries risk levels (91.7%), demonstrating superior performance in basic risk categorization compared to other models. Similarly, in the study conducted by Keleş and Arslan on dental trauma, Gemini gave the best result among chatbots with a 100% accuracy rate in diagnosis.[\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e] In the study conducted by G\u0026uuml;ven and colleagues on trauma, it was stated that Gemini provided more accurate and comprehensive answers also.[\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]\u003c/p\u003e \u003cp\u003eOn the other hand, a pilot study on pediatric dentistry indicated that chatbots have a lower accuracy rate in diagnosis than dentists, and while they may be useful for education and patient information, they cannot replace clinicians in diagnostic decision-making processes.[\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e] Besides, correct risk categorization alone does not ensure appropriate patient management. In pediatric dentistry, caries risk assessment requires translating risk indicators into individualized, age-appropriate management strategies, a process that remains highly dependent on clinician expertise and contextual judgment.[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]\u003c/p\u003e \u003cp\u003eIn the current study, significant differences were detected among AI models regarding clinical and radiographic follow-up recommendations (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.013; Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). Specifically, Gemini demonstrated the highest alignment with AAPD guidelines, achieving a 'correct/complete' response rate of 55.6%, whereas ChatGPT Plus showed the lowest performance at 19.4%. This striking disparity suggests that the models' utility depends not merely on static data retrieval but on their capacity to interpret dynamic guideline tables that integrate a patient\u0026rsquo;s age, risk category, and social determinants. AAPD follow-up protocols prescribe specific recall intervals and radiographic frequencies based on risk levels (low, moderate, high), requiring a level of contextual reasoning more complex than simple categorical diagnosis.[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e] The underperformance of otherwise advanced models like ChatGPT Plus in this domain may stem from insufficient specialization within the training data regarding preventive dentistry protocols or limitations in algorithmic weighting of clinical priorities. From a clinical perspective, inconsistent follow-up recommendations are concerning, as they could lead to either the progression of undetected early-stage lesions or unnecessary radiation exposure for pediatric patients, further reinforcing that chatbot outputs must remain subject to professional oversight.\u003c/p\u003e \u003cp\u003ePreventive intervention recommendations showed the greatest heterogeneity among the evaluated AI models. In the domains of fluoride therapy and dietary counseling, Claude significantly outperformed all other chatbots (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.010; \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001). While Gemini and Copilot showed higher accuracy in fissure sealant recommendations (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.006), they provided mostly partially correct advice for dietary and fluoride protocols. The high performance of Claude in dietary counseling (66.6% correct/complete) suggests a more robust integration of the AAPD Policy on Dietary Recommendations compared to its peers.[\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e] Conversely, the higher accuracy observed for Gemini and Copilot in fissure sealant recommendations (both 80.6% correct/complete) may reflect the relatively straightforward and clearly defined nature of AAPD guidance in this area, in contrast to dietary and fluoride recommendations, which are broader and require more contextual interpretation. These findings are consistent with broader reviews of healthcare chatbots, which often report that AI systems deliver generalized preventive advice and fail to tailor recommendations according to individual risk profiles.[\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e] To our knowledge, evidence on chatbot performance regarding fissure sealant and dietary recommendations is lacking; however, a prior study evaluating fluoride-related responses reported higher accuracy for Gemini compared with other models, differing from the results of the present study.[\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e] In another study conducted by Karam\u0026uuml;ft\u0026uuml;oğlu et al. regarding fluoride use, ChatGPT-4.0 was found to be superior to the Gemini model.[\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e] The overall inconsistency across preventive categories underscores that AI systems often struggle to bridge the gap between risk identification and the formulation of a comprehensive, evidence-based management plan.\u003c/p\u003e \u003cp\u003eAccording to the study results, no statistically significant difference was observed among the five AI models regarding restorative treatment approaches (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.480; Table\u0026nbsp;\u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e5\u003c/span\u003e). While Claude exhibited the highest \"correct/complete\" response rate at 66.7%, the success rates of the other models ranged between 47.2% and 55.6%. These findings align with the study conducted by Ozdemir and Yapici, which reported comparable performance across several large language models in responding to dentistry-specific restorative questions. [\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]\u003c/p\u003e \u003cp\u003eThis consistency across models may be attributed to the well-established and prescriptive nature of restorative protocols for cavitated or dentin-level lesions, which leave less room for interpretation compared to preventive strategies. However, the fact that none of the models achieved 100% complete success, and that a significant portion of responses were only \"partially correct\" (ranging from 33.3% to 52.8%), proves that AI still requires professional supervision in complex clinical scenarios. In pediatric dentistry, the restorative decision-making process is dynamic, shaped not only by the presence of a lesion but also by the child's age, level of cooperation, and the anticipated exfoliation time of the tooth. Consequently, while AI models may serve as helpful auxiliary tools in restorative planning, they are not yet reliable enough to serve as autonomous decision-makers in place of clinical judgment.\u003c/p\u003e \u003cp\u003eA significant finding of this study relates to the temporal consistency of the AI models' responses. While most models generally produced stable outputs across the three separate days of evaluation, Microsoft Copilot demonstrated a statistically significant level of daily variability (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.023; Table\u0026nbsp;\u003cspan refid=\"Tab6\" class=\"InternalRef\"\u003e6\u003c/span\u003e). Consistent with this finding, Barbosa et al. reported that Copilot exhibited the lowest consistency in their study as well.[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e] This indicates that the model may provide divergent clinical recommendations over time, even when presented with identical cases.\u003c/p\u003e \u003cp\u003eSuch variability poses a serious challenge to the reliability and reproducibility of clinical decision-support mechanisms. The fact that most evaluated systems are closed-source and undergo continuous updates may be the primary reason for these temporal fluctuations. As emphasized in the literature, concerns regarding the transparency and consistency of AI-generated medical advice remain highly relevant.[\u003cspan additionalcitationids=\"CR28\" citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e] In a clinical setting, recommending different risk categories or follow-up intervals for the same patient profile on different days could undermine standardization and jeopardize patient safety. Consequently, the fluctuations exhibited by models like Copilot reaffirm that AI tools should not yet function as standalone decision-makers but should instead be utilized as auxiliary tools under professional supervision.\u003c/p\u003e \u003cp\u003eThis study is significant as one of the few investigations in pediatric dentistry and, to our knowledge, the first to systematically compare five contemporary LLMs specifically for caries risk assessment. A major methodological strength is the use of 12 standardized AAPD-based cases evaluated over three separate days, resulting in a robust dataset of 1,080 records. Furthermore, adherence to STROBE and CHART guidelines ensures high transparency and rigor. By assessing multi-dimensional clinical decision-making, from follow-up to restorative planning, this research offers a comprehensive analysis directly relevant to pediatric dental practice.\u003c/p\u003e \u003cp\u003eSeveral limitations should be considered when interpreting these findings. While the use of standardized simulated cases ensured methodological control and reproducibility, it may not fully capture the complexity of real-world clinical practice, such as behavioral variability and parental factors. Additionally, since the chatbots were explicitly instructed to follow the AAPD guidelines, their performance under alternative frameworks like CAMBRA or Cariogram remained outside the scope of this study rather than being a functional deficiency. Finally, as these are closed-source systems subject to continuous updates, the observed temporal variability indicates that the clinical reliability of these tools may fluctuate over time, reflecting a challenge inherent to current generative AI technology.\u003c/p\u003e"},{"header":"CONCLUSION","content":"\u003cp\u003e In conclusion, while AI chatbots demonstrate promising capability in basic pediatric caries risk categorization, significant limitations remain in translating risk status into guideline-adherent follow-up schedules and preventive strategies. Although these systems may support education and preliminary clinical decision-making, they currently lack the consistency and clinical nuance required for autonomous use in pediatric caries risk assessment and management. Until further refinement and real-world clinical validation are achieved, AI chatbots should be regarded as complementary tools rather than substitutes for professional judgment in pediatric dentistry.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e: Not applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication:\u003c/strong\u003e Not applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability:\u0026nbsp;\u003c/strong\u003eThe datasets analyzed during the current study are available from the corresponding author upon reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests:\u003c/strong\u003e The authors declare no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding:\u003c/strong\u003e No funding.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors\u0026rsquo; contributions:\u003c/strong\u003e E.C.T. conceived the ideas; E.C.T. and Z.B.A. collected the data; Z.B.A. analysed the data; E.C.T. and Z.B.A. led the writing of the manuscript. All authors read and approved the final manuscript.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements:\u003c/strong\u003e None.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eWorld Health Organization. Global oral health status report: towards universal health coverage for oral health by 2030. World Health Organization; 2022.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFejerskov O, Nyvad B, Kidd E. Dental caries: the disease and its clinical management. Wiley; 2015.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAmerican Academy of Pediatric Dentistry. Caries-risk assessment and management for infants, children, and adolescents. The Reference Manual of Pediatric Dentistry. Chicago, IL: American Academy of Pediatric Dentistry 325\u0026thinsp;\u0026ndash;\u0026thinsp;31; 2025.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAmerican Dental Association on behalf of the Dental QualityAlliance. Guidance on caries risk assessment in children. 2018.\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ada.org/~/media/ADA/DQA/CRA_Report.pd-f?la=en\u003c/span\u003e\u003cspan address=\"https://www.ada.org/~/media/ADA/DQA/CRA_Report.pd-f?la=en\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Accessed 4 March 2026.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIssrani R, Alnusayri SAF, Alderaan DAA, Alruwaili SR, Almufarrij RAS, Alkhershawy LHS. Risk Factors for Caries in Children and Adolescents: A Systematic Review. Open Dentistry J. 2025;19:1\u0026ndash;13.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNg TC-H, Luo BW, Lam WY-H, Baysan A, Chu C-H, Yu OY. Updates on caries risk assessment\u0026mdash;a literature review. Dentistry J. 2024;12:312.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRechmann P, Chaffee B, Rechmann B, Featherstone J. Changes in caries risk in a practice-based randomized controlled trial. Adv Dent Res. 2018;29:15\u0026ndash;23.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFeatherstone JD, Alston P, Chaffee BW, Rechmann P. Caries Management by Risk Assessment (CAMBRA): an update for use in clinical practice for patients aged 6 through adult. J Calif Dent Assoc. 2019;47:25\u0026ndash;34.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBratthall D, H\u0026auml;nsel Petersson G. Cariogram\u0026ndash;a multifactorial risk assessment model for a multifactorial disease. Community Dent Oral Epidemiol. 2005;33:256\u0026ndash;64. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1111/j.1600-0528.2005.00233.x\u003c/span\u003e\u003cspan address=\"10.1111/j.1600-0528.2005.00233.x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSandmann S, Riepenhausen S, Plagwitz L, Varghese J. Systematic analysis of ChatGPT, Google search and Llama 2 for clinical decision support tasks. Nat Commun. 2024;15(1):2050.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDeng J, Heybati K, Park Y-J, Zhou F, Bozzo A. Artificial intelligence in clinical practice: a look at ChatGPT. Cleve Clin J Med. 2024;91:173\u0026ndash;80.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBarbosa IA, Alves MSA, de Almeida PRZ, de Almeida Rodrigues P, de Oliveira RP, de Menezes SAF, de Moura JDM, de Souza Fonseca RR. (2025) Assessing the diagnostic and treatment accuracy of large Language models (LLMs) in Peri-Implant diseases: a clinical experimental study. J Dent:106091.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKeleş \u0026Ouml;K, Arslan ZB. Performance of artificial intelligence chatbots in the diagnosis and management of simulated dental trauma cases: an evaluation based on IADT guidelines. Clin Oral Invest. 2026;30:26.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ede Moura JDM, Fontana CE, da Silva Lima VHR, de Souza Alves I, de Melo Santos PA, de Almeida Rodrigues P. Comparative accuracy of artificial intelligence chatbots in pulpal and periradicular diagnosis: A cross-sectional study. Comput Biol Med. 2024;183:109332.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMustuloğlu Ş, Deniz BP. Evaluation of chatbots in the emergency management of avulsion injuries. Dent Traumatol. 2025;41(4):437\u0026ndash;44.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBayraktar Nahir C. Can ChatGPT be guide in pediatric dentistry? BMC Oral Health. 2025;25:9. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/s12903-024-05393-1\u003c/span\u003e\u003cspan address=\"10.1186/s12903-024-05393-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKaya İ, Demirel A. Evaluation of Accuracy, Information Quality, and Readability of Artificial Intelligence Based Chatbots in Pediatric Oral Surgery: A Comparative Analysis Based on the AAPD Clinical Guideline. Cumhuriyet Dent J. 2025;28:586\u0026ndash;93.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMalta M, Cardoso LO, Bastos FI, Magnanini MMF, Silva, CMFPd. STROBE initiative: guidelines on reporting observational studies. Rev Saude Publica. 2010;44:559\u0026ndash;65.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHuo B, Collins GS, Chartash D, Thirunavukarasu AJ, Flanagin A, Iorio A, Cacciamani G, Chen X, Liu N, Mathur P. Reporting guideline for chatbot health advice studies: the CHART statement. JAMA Netw open. 2025;8:e2530220\u0026ndash;2530220.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAmerican Academy of Pediatric Dentistry. Policy on dietary recommendations for infants, children, and adolescents. The Reference Manual of Pediatric Dentistry. Chicago, IL: American Academy of Pediatric Dentistry 118\u0026thinsp;\u0026ndash;\u0026thinsp;22; 2025.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGuven Y, Ozdemir OT, Kavan MY. Performance of Artificial Intelligence Chatbots in Responding to Patient Queries Related to Traumatic Dental Injuries: A Comparative Study. Dent Traumatol. 2025;41:338\u0026ndash;47. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1111/edt.13020\u003c/span\u003e\u003cspan address=\"10.1111/edt.13020\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRokhshad R, Zhang P, Mohammad-Rahimi H, Pitchika V, Entezari N, Schwendicke F. Accuracy and consistency of chatbots versus clinicians for answering pediatric dentistry questions: A pilot study. J Dent. 2024;144:104938.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBiswas S, Biswas S. (2023) Title: Role of ChatGPT in Dental Science. Available at SSRN 4403581, 2023. Accessed 4 March 2026.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAksu S, Bakır FA. Yapay zeka sohbet robotlarının \u0026lsquo;diş hekimliğinde flor\u0026rsquo;konusu ile ilgili sorulara verdikleri yanıtların değerlendirilmesi. Mersin \u0026Uuml;niversitesi Tıp Fak\u0026uuml;ltesi Lokman Hekim Tıp. Tarihi ve Folklorik Tıp Dergisi. 2025;15:1069\u0026ndash;78.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKaram\u0026uuml;ft\u0026uuml;oğlu N, Varol EA, Bal C. Exploring artificial intelligence chatbots in pediatric fluoride education: a cross-sectional study. Sci Rep. 2025;16:182. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41598-025-28857-y\u003c/span\u003e\u003cspan address=\"10.1038/s41598-025-28857-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOzdemir ZM, Yapici E. Evaluating the accuracy, reliability, consistency, and readability of different large language models in restorative dentistry. J Esthetic Restor Dentistry. 2025;37:1740\u0026ndash;52.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMakrygiannakis MA, Giannakopoulos K, Kaklamanos EG. Evidence-based potential of generative artificial intelligence large language models in orthodontics: a comparative study of ChatGPT, Google Bard, and Microsoft Bing. Eur J Orthod. 2026;48:cjae017.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDermata A, Arhakis A, Makrygiannakis MA, Giannakopoulos K, Kaklamanos EG. Evaluating the evidence-based potential of six large language models in paediatric dentistry: a comparative study on generative artificial intelligence. Eur Archives Pediatr Dentistry. 2025;26:527\u0026ndash;35.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlhaidry HM, Fatani B, Alrayes JO, Almana AM, Alfhaed NK, Alhaidry H, Alrayes J, Almana A and Alfhaed, Sr NK. (2023) ChatGPT in dentistry: a comprehensive review. Cureus 15(4): e38317.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-oral-health","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"ohea","sideBox":"Learn more about [BMC Oral Health](http://bmcoralhealth.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/ohea/default.aspx","title":"BMC Oral Health","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"artificial intelligence, chatbots, caries risk assessment, pediatric dentistry, clinical decision-making, AAPD guidelines","lastPublishedDoi":"10.21203/rs.3.rs-9070138/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9070138/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eObjectives\u003c/h2\u003e \u003cp\u003e This study evaluated the accuracy and guideline alignment of artificial intelligence (AI)\u0026ndash;based chatbots in pediatric caries risk assessment and management by comparing their recommendations with the American Academy of Pediatric Dentistry (AAPD) caries risk assessment guideline using simulated pediatric cases.\u003c/p\u003e\u003ch2\u003eMaterials and Methods\u003c/h2\u003e \u003cp\u003e A case-based comparative study was conducted using 12 simulated pediatric patient profiles representing low, moderate, and high caries risk categories constructed according to the AAPD guideline. Five AI chatbots\u0026mdash;ChatGPT-5.2 (Plus), ChatGPT-4o (Free), Microsoft Copilot, Google Gemini 1.5, and Claude\u0026mdash;were evaluated. A standardized prompt instructed each chatbot to assess caries risk, recommend clinical and radiographic follow-up frequency, propose preventive interventions, and outline restorative approaches based on the AAPD guideline. Each case was presented to each chatbot on three separate days, generating 1,080 responses. Outputs were evaluated using a guideline-based scoring rubric by an experienced pediatric dentist. Statistical analyses included chi-square tests for between-chatbot comparisons and Friedman tests for within-model consistency (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05).\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eNo significant differences were observed among the chatbots in caries risk classification (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.059). However, significant differences were found in clinical follow-up recommendations (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.013) and preventive interventions. Claude demonstrated higher accuracy in dietary counseling and fluoride therapy (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.001; \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.010), while Gemini and Copilot performed better in fissure sealant recommendations (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.006). No differences were observed in restorative treatment recommendations (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.480).\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e \u003cp\u003e AI chatbots were generally capable of identifying pediatric caries risk levels; however, inconsistencies were observed when translating risk status into guideline-based follow-up and preventive recommendations.\u003c/p\u003e\u003ch2\u003eClinical Relevance:\u003c/h2\u003e \u003cp\u003eAI chatbots may support dental education and preliminary clinical decision-making in pediatric dentistry, but their recommendations should be interpreted cautiously and cannot replace professional clinical judgment.\u003c/p\u003e","manuscriptTitle":"Artificial Intelligence in Pediatric Dentistry: Are Chatbots Aligned With AAPD Caries Risk Assessment Guideline?","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-19 16:48:26","doi":"10.21203/rs.3.rs-9070138/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"editorInvitedReview","content":"","date":"2026-04-20T18:58:53+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"50492168947150006814619050444446215082","date":"2026-04-08T14:10:35+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-03-17T12:34:50+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2026-03-12T04:50:18+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-03-12T03:24:17+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-03-12T03:23:44+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Oral Health","date":"2026-03-09T08:05:29+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-oral-health","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"ohea","sideBox":"Learn more about [BMC Oral Health](http://bmcoralhealth.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/ohea/default.aspx","title":"BMC Oral Health","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"d1656103-5e0e-41fe-b33c-ec67dc8fc3c1","owner":[],"postedDate":"March 19th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-03-19T16:48:26+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-19 16:48:26","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9070138","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9070138","identity":"rs-9070138","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}