Abstract
Aim
This study aimed to evaluate and compare the knowledge of three artificial intelligence (AI) systems and three dental specialist groups regarding the management of traumatic dental injuries (TDIs).
Methodology
An online survey consisting of three sections was administered: (1) an information sheet with electronic informed consent, (2) five sociodemographic questions, and (3) thirty multiple-choice questions assessing TDI knowledge based on the International Association of Dental Traumatology (IADT) guidelines and current literature. The survey was distributed via e-mail to specialists in endodontics, pediatric dentistry, and oral and maxillofacial surgery through the Turkish Dental Association database.
For the AI systems (ChatGPT-4o, Gemini Advanced, and DeepSeek-V3), the 30-item questionnaire was administered under a standardized prompt three times per day over 10 consecutive days, and all responses were recorded. Internal consistency and reliability were assessed using Cronbach’s alpha and intraclass correlation coefficients (ICC). Statistical analyses were performed using one-way ANOVA and independent-samples t-tests (p < 0.05).
Results
A total of 95 specialists participated in the study (64 females, 31 males). Female participants demonstrated significantly higher correct-response rates than males. Participants classified as having very good knowledge achieved significantly higher scores than those with acceptable or low knowledge levels. No statistically significant differences were observed according to professional experience or workplace setting.
Significant differences were found among specialist groups, with pediatric dentistry specialists achieving the highest correct-response rate. The overall mean accuracy among all specialists was 70.98% (SD = 15.6). AI systems performed as follows: ChatGPT-4o, 86% (SD = 3.2); Gemini Advanced, 79.33% (SD = 3.75); and DeepSeek-V3, 72.67% (SD = 4.32).
Conclusions
AI systems demonstrated higher overall accuracy than human participants; however, certain items resulted in misleading responses, highlighting the importance of case-specific interpretation. AI tools may serve as supportive adjuncts in clinical decision-making, but further rigorously designed studies are required to clarify their optimal role and limitations in TDI management.
Supplementary Information
The online version contains supplementary material available at 10.1186/s12903-026-08106-y.
Keywords: Artificial intelligence, Dental specialists, Knowledge assessment, Large language models, Traumatic dental injuries
Introduction
Traumatic dental injuries (TDIs) are sudden impact injuries involving the teeth and surrounding supporting structures, and they frequently necessitate urgent clinical intervention [1]. These injuries are particularly prevalent among children and young adults, accounting for approximately 5% of all types of bodily injuries [2]. Earlier reports suggested that TDIs could emerge as one of the most significant oral health problems affecting younger populations [3], and subsequent investigations have demonstrated that their prevalence remains considerable, in some populations even surpassing that of dental caries and periodontal diseases [1].
Beyond the immediate physical consequences, TDIs exert a substantial negative influence on quality of life. Children presenting with fractured teeth have been shown to experience greater difficulties with eating, oral hygiene, and psychosocial functioning compared with those without a history of dental trauma [4]. Consequently, the effective management of TDIs requires clinicians to possess comprehensive theoretical knowledge and advanced clinical competence across all stages of trauma care [5]. Nevertheless, studies from various countries consistently indicate that dentists’ knowledge regarding the appropriate management of TDIs is frequently inadequate [5–7].
Artificial intelligence (AI) represents a field of computer science focused on the development of systems capable of simulating human cognitive processes [8]. Large language models (LLMs), commonly referred to as AI-driven chatbots, have recently emerged as an important paradigm within AI research and application [9]. Growing evidence suggests that LLMs may support both medical education and clinical decision-making [10, 11]; however, their reliability in complex, context-dependent clinical scenarios remains uncertain [12]. Given that TDIs management often requires rapid, accurate, and frequently multidisciplinary decision-making, evaluating the potential role of AI systems in supporting expert judgment is of increasing relevance.
Accordingly, the present study sought to compare the knowledge levels of three artificial intelligence models (ChatGPT-4o, Gemini Advanced, and DeepSeek-V3) with those of specialists representing three dental disciplines (Endodontics, Pediatric Dentistry, and Oral and Maxillofacial Surgery) regarding the management of TDIs. The null hypothesis was that no significant differences would exist between the groups.
Materials and methods
For the sample-size calculation, a difference corresponding to an effect size of f = 0.40 was assumed. Accordingly, the a priori power analysis indicated that 28 specialists per group (total = 84 participants) would be required to achieve 90% statistical power at α = 0.05. The study protocol received approval from the Hacettepe University Health Sciences Research Ethics Committee (Approval No: SBA 24/1205). The objectives, scope, and confidentiality procedures of the study were explained in the opening section of the online questionnaire, and electronic informed consent was obtained from all participants prior to enrollment.
Participants were eligible if they had graduated from a national dental school, possessed a specialty degree in Endodontics, Pediatric Dentistry, or Oral and Maxillofacial Surgery, and were actively practicing in a university hospital, private dental hospital, public hospital, oral and dental health center, or private clinic. Dental students, interns, dentists who had not yet completed specialty training, individuals with specialties outside the three target disciplines, and general dentists without specialty certification were excluded. Questionnaires that were incomplete were also omitted from the final analysis.
The questionnaire was designed to assess dentists’ knowledge regarding TDIs. Items were developed based on the guidelines of the International Association of Dental Traumatology (IADT) [2, 13–15] and supported by current literature [16–18] .
A structured, multi-step approach was adopted to ensure content validity and clinical relevance. An initial pool of items was generated and subsequently reviewed by two independent subject-matter experts (a senior endodontist and a pediatric dentist), who evaluated clarity, clinical accuracy, and alignment with contemporary guideline recommendations. Revisions were made to eliminate ambiguity and improve representativeness of real-world TDI scenarios.
The final instrument was structured around clinically relevant TDI domains representing frequently encountered and decision-critical injury patterns in practice, including luxation injuries, crown and root fractures, avulsion injuries, and associated periodontal trauma. Both knowledge-based and case-based formats were incorporated to assess factual recall as well as applied clinical reasoning under standardized conditions. Internal consistency of the 30-item questionnaire was evaluated using Cronbach’s alpha coefficient.
The final questionnaire was created in Microsoft Word (Microsoft Corp., Redmond, USA) and administered online via Google Forms (Google LLC, Mountain View, USA). It comprised three sections:
Sect. 1: Participant information sheet and electronic informed consent.
Sect. 2: Five items assessing socio-demographic characteristics (gender, specialty field, years of professional experience, workplace setting, and self-reported knowledge regarding TDIs, assessed through participants’ self-classification into one of four predefined categories: low, acceptable, good, or very good).
Sect. 3: Thirty multiple-choice items evaluating knowledge related to TDIs. Topics included crown and crown–root fractures (Q1–Q6), root fractures (Q7–Q10), alveolar fractures (Q11), periodontal support injuries (Q12–Q13), extrusive luxation (Q14), intrusive luxation (Q15–Q16), lateral luxation (Q17–Q18), and avulsion injuries (Q19–Q30).
The complete questionnaire is provided as Supplementary file 1. The 30 multiple-choice questions consisted of 13 case-based and 17 knowledge-based items. The same questionnaire was also presented to the AI systems (ChatGPT-4o, Gemini Advanced, and DeepSeek-V3). Each AI model was queried three times per day over 10 consecutive days, resulting in 30 independent response sets per system. All assessments were conducted in newly initiated, independent chat sessions to prevent contextual carryover between queries. Chain-of-Thought (CoT) prompting was not employed. The AI models were instructed, using a standardized prompt, to select a single correct answer for each multiple-choice question. No step-by-step reasoning or justification was requested, and only the selected option (A–E) was recorded for analysis. For each question, the models were instructed to provide only one answer using a standardized prompt: “Please select only one correct answer for the following multiple-choice questions.” The order of the questions was kept constant across all sessions to ensure methodological consistency and comparability between models. No investigator-driven updates, fine-tuning, or parameter modifications were applied during the data collection period. The survey was distributed to specialists in Endodontics, Pediatric Dentistry, and Oral and Maxillofacial Surgery via email through the Turkish Dental Association database.
Statistical analysis
Responses were compiled in Microsoft Excel (Microsoft Corp., Redmond, USA) and analyzed using IBM SPSS Statistics, version 23.0 (IBM Corp., Armonk, NY, USA). Internal consistency and reliability of the questionnaire were assessed using Cronbach’s alpha, while the response consistency of the artificial intelligence systems across repeated administrations was evaluated using intraclass correlation coefficients (ICC) and Cronbach’s alpha. Descriptive statistics (frequency, percentage, mean, and standard deviation) were calculated. Differences in mean numbers of correct responses among groups were examined using One-Way ANOVA and independent-samples t-tests For comparisons involving repeated measurements obtained from artificial intelligence systems across multiple administrations, repeated-measures analysis of variance (repeated-measures ANOVA) was performed. A significance level of p < 0.05 was considered statistically significant.
Results
The survey was distributed via email using the Dental Association database, and a total of 105 dentists provided complete responses. Ten individuals did not meet the inclusion criteria and were excluded; therefore, the responses of 95 participants were included in the analysis. Of the participating specialists, 30 were endodontists, 33 were pediatric dentists, and 32 were oral and maxillofacial surgeons. The sociodemographic characteristics of the study population are summarized in Table 1, and the distribution of correct responses is presented in Tables 2, 3, 4, 5 and 6.
Table 1.
Distribution of participants according to their sociodemographic characteristics
| N | % | |
|---|---|---|
| Gender | ||
| Female | 64 | 67.4 |
| Male | 31 | 32.6 |
| Professional experience (years) | ||
| ≤ 5 | 30 | 31.6 |
| 6–10 | 45 | 47.4 |
| 11–15 | 15 | 15.8 |
| ≥ 16 | 5 | 5.3 |
| Specialty | ||
| Endodontics | 30 | 31.6 |
| Pediatric Dentistry | 33 | 34.7 |
| Oral and Maxillofacial Surgery | 32 | 33.7 |
| Place of employment | ||
| University | 34 | 35.8 |
| Public | 30 | 31.6 |
| Private | 31 | 32.6 |
| Self-reported TDI knowledge level | ||
| Low | 2 | 2.1 |
| Acceptable | 24 | 25.3 |
| Good | 48 | 50.5 |
| Very Good | 21 | 22.1 |
Table 2.
Correct answer rates according to gender
| Mean ± SD (accuracy proportion) | Min – Max | p-Value | |
|---|---|---|---|
| Gender | |||
| Female | 0.75 ± 0.13 | 0.27–0.97 | < 0.001* |
| Male | 0.63 ± 0.17 | 0.1–0.97 | |
Independent samples t-test. *p < 0.05 indicates statistical significance
Table 3.
Correct answer rates according to professional experience
| Mean ± SD (accuracy proportion) | Min-Max | p-Value | |
|---|---|---|---|
| Professional experience (years) | |||
| ≤ 5 | 0.74 ± 0.15 | 0.27–0.97 | 0.118 |
| 6–10 | 0.71 ± 0.17 | 0.10–0.97 | |
| 11–15 | 0.68 ± 0.11 | 0.50–0.90 | |
| ≥ 16 | 0.57 ± 0.15 | 0.43–0.77 | |
One-way ANOVA
Table 4.
Correct answer accuracy rates according to place of employment
| Mean ± SD (accuracy proportion) | Min-Max | p-Value | |
|---|---|---|---|
| Place of employment | |||
| Public | 0.71 ± 0.14 | 0.27–0.93 | 0.987 |
| Private | 0.71 ± 0.18 | 0.10–0.97 | |
| University | 0.71 ± 0.14 | 0.27–0.97 | |
One-way ANOVA
Table 5.
Correct answer accuracy rates according to self-reported knowledge level on TDIs
| Mean ± SD (accuracy proportion) | Min-Max | p-Value | |
|---|---|---|---|
| Self-reported TDI knowledge level | |||
| Very Good | 0.76 ± 0.11 | 0.53–0.97 | 0.025 |
| Good | 0.72 ± 0.15 | 0.43–0.97 | |
| Acceptable -Low | 0.64 ± 0.18 | 0.10–0.90 | |
One-way ANOVA
Table 6.
Correct answer accuracy rates according to dental specialty
| Mean ± SD (accuracy proportion) | Min-Max | p-Value | |
|---|---|---|---|
| Dental specialties | |||
| Oral and Maxillofacial Surgery | 0.64 ± 0.11 | 0.27–0.80 | < 0.001* |
| Endodontics | 0.69 ± 0.20 | 0.10–0.97 | |
| Pediatric Dentistry | 0.80 ± 0.10 | 0.57–0.97 | |
One-way ANOVA
A significant difference was observed between genders (p < 0.001), with females demonstrating higher overall scores. One-way ANOVA also revealed statistically significant differences across self-reported knowledge levels and specialties (p < 0.05).
The overall correct response rate among all participants was 70.98% (SD = 15.6). When grouped by specialty, Pediatric Dentistry specialists achieved the highest proportion of correct responses (80.10%, SD = 9.94), followed by Endodontics (68.89%, SD = 19.85) and Oral and Maxillofacial Surgery (63.54%, SD = 10.84). The percentage of correct responses for each individual question is shown in Fig. 1.
Fig. 1.
Percentage of correct responses given by participants for each question
Across the artificial intelligence systems, the mean correct response rates over the 10-day evaluation period were 86% (SD = 3.2) for ChatGPT-4o, 79.33% (SD = 3.75) for Gemini Advanced, and 72.67% (SD = 4.32) for DeepSeek-V3. Comparative accuracy outcomes for specialists and AI models are illustrated in Fig. 2.
Fig. 2.
Overall correct answer rates (%) of dental specialists and artificial intelligence systems
Cronbach’s alpha coefficients for the AI systems were 0.992 for ChatGPT-4o, 0.988 for Gemini Advanced, and 0.987 for DeepSeek-V3. The intraclass correlation coefficients (ICC) were 0.808, 0.737, and 0.710, respectively. Trends in correct responses generated by the AI models across the 10-day period are presented in Fig. 3.
Fig. 3.
Change in the total number of correct responses provided by the AI-powered chatbots over the 10-day period
According to question type, correct response rates varied significantly across dental specialties and AI systems. Pediatric dentistry specialists and ChatGPT-4o demonstrated higher accuracy than the other groups across both case-based and knowledge-based questions, as shown Table 7.
Table 7.
Correct response rates according to question type across dental specialties and artificial intelligence systems
| Group | Case-based questions (Mean ± SD) | Knowledge-based questions (Mean ± SD) |
|---|---|---|
| Dental specialties | ||
| Oral and Maxillofacial Surgery | 0.58 ± 0.15 | 0.68 ± 0.15 |
| Endodontics | 0.63 ± 0.20 | 0.74 ± 0.21 |
| Pediatric Dentistry | 0.72 ± 0.14 | 0.87 ± 0.10 |
| p-value | 0.005 | < 0.001* |
| Artificial intelligence systems | ||
| ChatGPT-4o | 0.77 ± 0.07 | 0.93 ± 0.02 |
| Gemini Advanced | 0.75 ± 0.05 | 0.83 ± 0.05 |
| DeepSeek-V3 | 0.67 ± 0.07 | 0.77 ± 0.06 |
| p-value | < 0.001* | < 0.001* |
Values are presented as mean correct response rates ± standard deviation (SD). Between-specialty comparisons were performed using one-way ANOVA, and comparisons among artificial intelligence systems using repeated-measures ANOVA. *p < 0.05 indicates statistical significance.
In addition, the internal consistency of the 30-item questionnaire was confirmed, with a Cronbach’s alpha value of 0.78, exceeding the recommended threshold of 0.70 [19], and demonstrating that the instrument exhibited acceptable reliability.
Discussion
The multidisciplinary nature of TDI management and the frequent need for urgent intervention require clinicians to formulate treatment decisions rapidly and accurately. Within endodontics in particular, the complexity and diversity of treatment options necessitate an evaluation of whether AI systems can provide accurate and clinically reliable guidance for trauma-related scenarios [20]. In the present study, knowledge regarding TDI management was compared across three dental specialty groups and three contemporary AI models. Significant performance differences were identified both among clinicians and across the AI systems, highlighting notable variability in decision-making processes between humans and AI tools.
According to the present research, the available literature has not yet simultaneously examined TDIs knowledge across three dental specialties and three artificial intelligence chatbots using a repeated and standardized evaluation protocol, suggesting that this combined human–AI comparison adds a meaningful contribution to the existing evidence base. A recent study compared the guideline-based knowledge of general dentists with that of ChatGPT-4o and Gemini, reporting higher accuracy for ChatGPT-4o across most items [21]. While the present findings are broadly consistent with this report, the inclusion of multiple dental specialties, a larger item set, and an additional contemporary AI model in the current study provides complementary and extended insight into human–AI performance in TDIs management. AI systems rely on probabilistic pattern recognition derived from training data rather than experiential clinical reasoning, which may explain their higher overall accuracy yet misleading responses in context-sensitive scenarios. These findings likely reflect differences in cognitive and training paradigms rather than simple superiority of one group over another. In addition, both clinicians and AI systems demonstrated higher accuracy in knowledge-based questions than in case-based scenarios. This pattern suggests that recalling guideline-based information may be less demanding than applying such knowledge within complex clinical decision-making contexts, highlighting a potential gap between theoretical knowledge and clinical application in TDI management.
The AI systems evaluated—ChatGPT, Gemini, and DeepSeek—were selected to represent widely used large language models developed using different architectures and training strategies. ChatGPT and Gemini have been most frequently assessed in dental trauma–related research [20, 22, 23], whereas DeepSeek was included as a newer model to explore performance variability across emerging AI platforms. Endodontics, pediatric dentistry, and oral and maxillofacial surgery specialists were included because TDI management constitutes a core component of both specialist training curricula and routine clinical practice in these disciplines.
The study included a greater proportion of female participants, consistent with several earlier reports [17, 24, 25]. This trend likely reflects national workforce data indicating that female dentists have outnumbered male dentists since the 1990s [26]. Although exploratory analyses suggested higher accuracy scores among women, this comparison was not a primary objective, and the unequal gender distribution limits interpretation. Moreover, previous studies reporting no gender-related differences [17, 27, 28] emphasize the need for more controlled investigations before drawing definitive conclusions.
Across specialties, pediatric dentistry specialists demonstrated significantly higher accuracy. This finding is plausible given the higher frequency of trauma cases involving children and the greater curricular emphasis on TDI management in pediatric training programs [17]. Nevertheless, the literature is not entirely consistent, as some studies have shown that endodontists attain higher knowledge levels [25], suggesting that differences in educational frameworks, clinical responsibilities, and patient populations may account for these discrepancies. The higher performance observed in pediatric dentistry may further reflect the routine integration of traumatic dental injury management into pediatric clinical practice, enabling more frequent reinforcement of guideline-based decision pathways. In contrast, oral and maxillofacial surgeons may encounter TDIs within broader trauma contexts, where management priorities extend beyond tooth-specific considerations. Beyond statistical significance, these differences may indicate fundamental distinctions in knowledge acquisition, clinical exposure, and decision-making frameworks, which likely shape how clinicians and AI systems approach trauma-related decisions.
Significant differences were also observed among AI models. ChatGPT-4o achieved the highest accuracy (86%), followed by Gemini Advanced (79.33%), whereas DeepSeek-V3 performed lowest (72.67%). This hierarchy aligns with findings from other medical contexts. Marcaccini et al. (2025) reported superior accuracy of ChatGPT-4o in hand fracture management compared with DeepSeek-V3 and Gemini [29]. Similarly, Seth et al. (2025) demonstrated high concordance for ChatGPT and Gemini in treatment recommendations for Dupuytren’s disease, with lower performance for DeepSeek [30]. A recent comparison within medical education revealed relatively similar outcomes among several models, though differences persisted [31]. Collectively, these results suggest that while AI systems show promising potential, their performance remains variable and context-dependent, and they cannot substitute for human clinical expertise [29–31].
The temporal evolution of AI capabilities is also noteworthy. Earlier investigations reported inferior performance compared with students [32], whereas more recent work suggests performance surpassing that of novice clinicians [12]. Furthermore, GPT-4o has demonstrated performance comparable to senior specialists in resuscitation assessments [33], reflecting rapid advancement across successive model iterations. These heterogeneous findings indicate that AI competence is dynamic and influenced by ongoing updates, training data, and fine-tuning processes.
LLMs may produce outputs that appear coherent yet lack factual accuracy, a phenomenon termed “hallucination” [34]. The multiple-choice format used in this study likely mitigated overt hallucinations, as only one response could be selected. However, temporal inconsistency across repeated prompts remains a limitation and warrants further systematic investigation.
Several limitations should be acknowledged. The sample comprised specialists from a single country, and performance was assessed using a 30-item questionnaire that may not fully reflect real-world clinical complexity. Visual diagnostic elements such as photographs or radiographs were not included. Additionally, AI models were tested using a single standardized prompt, and the influence of alternative prompting strategies was not examined. Despite these constraints, the comparison of three contemporary AI systems evaluated over multiple days, alongside different specialist groups, represents a notable methodological strength.
From an educational perspective, integrating AI-assisted learning tools with specialty-specific clinical training may help standardize trauma management knowledge and reduce variability in decision-making. Future research should therefore investigate hybrid human–AI collaborative models and evaluate how AI-supported educational interventions influence long-term clinical competence in TDI management.
Conclusion
Pediatric dentistry specialists exhibited comparatively higher levels of knowledge regarding the management of traumatic dental injuries. Among the AI systems evaluated, ChatGPT-4o demonstrated the greatest accuracy, surpassing the overall mean performance of the specialist groups. These findings indicate that large language models may serve as valuable adjuncts to clinical decision-making. Nevertheless, issues related to model reliability, contextual variability, and the inherent complexity of real-world trauma scenarios necessitate cautious interpretation. Further rigorously designed, clinically oriented investigations are required to more clearly delineate the appropriate scope and limitations of AI integration into routine dental practice.
Supplementary Information
Acknowledgements
The authors thank all specialist participants for their voluntary contributions, the Turkish Dental Association for supporting the dissemination of the survey, and Dr. Sevilay Karahan for her contribution to the statistical analyses.
Clinical trial number
Not applicable
Abbreviations
- AI
Artificial Intelligence
- IADT
International Association of Dental Traumatology
- ICC
Intraclass Correlation Coefficients
- LLMs
Large Language Models
- TDIs
Traumatic Dental Injuries
Authors’ contributions
Conceptualization: G.Z.S., A.E.; Methodology: G.Z.S., A.E.; Data acquisition: G.Z.S.; Writing – original draft: G.Z.S.; Writing – review & editing: G.Z.S., A.E.; Supervision: A.E.
Funding
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request.
Declarations
Ethics approval and consent to participate
The study was approved by the Hacettepe University Health Sciences Research Ethics Committee (Approval No: SBA 24/1205). The study was conducted in accordance with the principles of the Declaration of Helsinki. Informed consent was obtained electronically from all participants prior to participation.
Consent for publication
Not applicable. This study does not contain any individual person’s data, images, or other identifying clinical information.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.Lam R. Epidemiology and outcomes of traumatic dental injuries: a review of the literature. Aust Dent J. 2016;61(Suppl 1):4–20. 10.1111/adj.12395. [DOI] [PubMed] [Google Scholar]
- 2.Levin L, Day PF, Hicks L, O’Connell A, Fouad AF, Bourguignon C, et al. International Association of Dental Traumatology guidelines for the management of traumatic dental injuries: General introduction. Dent Traumatol. 2020;36(4):309–13. 10.1111/edt.12574. [DOI] [PubMed] [Google Scholar]
- 3.Andreasen JO, Andreasen FM. Dental traumatology: quo vadis. Dent Traumatol. 1990;6(2):78–80. 10.1111/j.1600-9657.1990.tb00395.x. [Google Scholar]
- 4.Ilma de Souza Cortes M, Marcenes W, Sheiham A. Impact of traumatic injuries to the permanent teeth on the oral health-related quality of life in 12–14-year-old children. Commun Dent Oral Epidemiol. 2002;30(3):193–8. 10.1034/j.1600-0528.2002.300305.x. [DOI] [PubMed] [Google Scholar]
- 5.Pedrini D, Panzarini SR, Poi WR, Sundefeld ML, Tiveron AR. Dentists’ level of knowledge of the treatment plans for periodontal ligament injuries after dentoalveolar trauma. Braz Oral Res. 2011;25(4):307–13. 10.1590/s1806-83242011000400005. [DOI] [PubMed] [Google Scholar]
- 6.Cauwels RG, Martens LC, Verbeeck RM. Educational background of Flemish dental practitioners and their perceptions of their management of dental trauma. Dent Traumatol. 2014;30(2):133–9. 10.1111/edt.12069. [DOI] [PubMed] [Google Scholar]
- 7.Kostopoulou MN, Duggal MS. A study into dentists’ knowledge of the treatment of traumatic injuries to young permanent incisors. Int J Paediatr Dent. 2005;15(1):10–9. 10.1111/j.1365-263X.2005.00588.x. [DOI] [PubMed] [Google Scholar]
- 8.Shan T, Tay FR, Gu L. Application of Artificial Intelligence in Dentistry. J Dent Res. 2021;100(3):232–44. 10.1177/0022034520969115. [DOI] [PubMed] [Google Scholar]
- 9.Günay S, Öztürk A, Yiğit Y. The accuracy of Gemini, GPT-4, and GPT-4o in ECG analysis: A comparison with cardiologists and emergency medicine specialists. Am J Emerg Med. 2024;84:68–73. 10.1016/j.ajem.2024.07.043. [DOI] [PubMed] [Google Scholar]
- 10.Liu M, Okuhara T, Dai Z, Huang W, Gu L, Okada H, et al. Evaluating the Effectiveness of advanced large language models in medical Knowledge: A Comparative study using Japanese national medical examination. Int J Med Inf. 2025;193:105673. 10.1016/j.ijmedinf.2024.105673. [DOI] [PubMed] [Google Scholar]
- 11.Kung TH, Cheatham M, Medenilla A, Sillos C, De Leon L, Elepaño C, et al. Performance of ChatGPT on USMLE: Potential for AI-assisted medical education using large language models. PLOS Digit Health. 2023;2(2):e0000198. 10.1371/journal.pdig.0000198. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Patel J, Robinson P, Illing E, Anthony B. Is ChatGPT 3.5 smarter than Otolaryngology trainees? A comparison study of board style exam questions. PLoS ONE. 2024;19(9):e0306233. 10.1371/journal.pone.0306233. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Bourguignon C, Cohenca N, Lauridsen E, Flores MT, O’Connell AC, Day PF, et al. International Association of Dental Traumatology guidelines for the management of traumatic dental injuries: 1. Fractures and luxations. Dent Traumatol. 2020;36(4):314–30. 10.1111/edt.12578. [DOI] [PubMed] [Google Scholar]
- 14.Fouad AF, Abbott PV, Tsilingaridis G, Cohenca N, Lauridsen E, Bourguignon C, et al. International Association of Dental Traumatology guidelines for the management of traumatic dental injuries: 2. Avulsion of permanent teeth. Dent Traumatol. 2020;36(4):331–42. 10.1111/edt.12573. [DOI] [PubMed] [Google Scholar]
- 15.Day PF, Flores MT, O’Connell AC, Abbott PV, Tsilingaridis G, Fouad AF, et al. International Association of Dental Traumatology guidelines for the management of traumatic dental injuries: 3. Injuries in the primary dentition. Dent Traumatol. 2020;36(4):343–59. 10.1111/edt.12576. [DOI] [PubMed] [Google Scholar]
- 16.Cvijic A, Åstrøm A, Kvernenes M, Tsilingaridis G, Bletsa A. Traumatic dental injuries: Knowledge assessment of dentists in the Norwegian Public Dental Service of Vestland. Dent Traumatol. 2024;40(4):398–409. 10.1111/edt.12945. [DOI] [PubMed] [Google Scholar]
- 17.Duruk G, Erel ZB. Assessment of Turkish dentists’ knowledge about managing avulsed teeth. Dent Traumatol. 2020;36(4):371–81. 10.1111/edt.12543. [DOI] [PubMed] [Google Scholar]
- 18.Buldur B, Kaptan A. Factors Associated with Knowledge and Attitude of Management of Traumatic Dental Injuries: A Cross-Sectional Study among Turkish Dentists. Pesquisa Brasileira em Odontopediatria e Clínica Integrada. 2018;18. 10.4034/PBOCI.2018.181.30.
- 19.Kilic S. Cronbachs Alpha Reliability Coefficient. J Mood Disorders. 2016;6:1. 10.5455/jmood.20160307122823. [Google Scholar]
- 20.Ozden I, Gokyar M, Ozden ME, Sazak Ovecoglu H. Assessment of artificial intelligence applications in responding to dental trauma. Dent Traumatol. 2024;40(6):722–9. 10.1111/edt.12965. [DOI] [PubMed] [Google Scholar]
- 21.Sağlam H, Sezgin GP, Kaplan T, Kaplan SS. Artificial intelligence chatbots versus dentists: a comparative knowledge assessment on traumatic dental injury management. BMC Oral Health. 2026. 10.1186/s12903-026-07728-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Tokgöz Kaplan T, Cankar M. Evidence-Based Potential of Generative Artificial Intelligence Large Language Models on Dental Avulsion: ChatGPT Versus Gemini. Dent Traumatol. 2025;41(2):178–86. 10.1111/edt.12999. [DOI] [PubMed] [Google Scholar]
- 23.Portilla ND, Garcia-Font M, Nagendrababu V, Abbott PV, Sanchez JAG, Abella F. Accuracy and Consistency of Gemini Responses Regarding the Management of Traumatized Permanent Teeth. Dent Traumatol. 2025;41(2):171–7. 10.1111/edt.13004. [DOI] [PubMed] [Google Scholar]
- 24.Tzanetakis GN, Tzimpoulas N, Markou M, Papanakou SI, Gizani S, Georgopoulou M. Evaluating the knowledge level, attitudes, and therapeutic approaches of Greek dentists for traumatic dental injuries. Dent Traumatol. 2021;37(2):177–87. 10.1111/edt.12623. [DOI] [PubMed] [Google Scholar]
- 25.Hartmann RC, Rossetti BR, Siqueira Pinheiro L, Poli de Figueiredo JA, Rossi-Fedele G, M SG, et al. Dentists’ knowledge of dental trauma based on the International Association of Dental Traumatology guidelines: A survey in South Brazil. Dent Traumatol. 2019;35(1):27–32. 10.1111/edt.12450. [DOI] [PubMed] [Google Scholar]
- 26.Yusufoğlu S, Tulumbacı F. Knowledge of general dentists, endodontists and pediatric dentists about the management on dental traumas: a survey in turkey. Brazilian Dent Sci. 2020;23. 10.14295/bds.2020.v23i3.2006.
- 27.Jadav NM, Abbott PV. Dentists’ knowledge of dental trauma based on the International Association of Dental Traumatology guidelines: An Australian survey. Dent Traumatol. 2022;38(5):374–80. 10.1111/edt.12761. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Bucchi C, Arroyo-Bote S. Knowledge and attitudes of dentists regarding traumatic dental injuries. Eur J Paediatr Dent. 2021;22(2):114–8. 10.23804/ejpd.2021.22.02.5. [DOI] [PubMed] [Google Scholar]
- 29.Marcaccini G, Seth I, Xie Y, Susini P, Pozzi M, Cuomo R, et al. Breaking Bones, Breaking Barriers: ChatGPT, DeepSeek, and Gemini in Hand Fracture Management. J Clin Med. 2025;14(6). 10.3390/jcm14061983. [DOI] [PMC free article] [PubMed]
- 30.Seth I, Marcaccini G, Lim K, Castrechini M, Cuomo R, Ng SK, et al. Management of Dupuytren’s Disease: A Multi-Centric Comparative Analysis Between Experienced Hand Surgeons Versus Artificial Intelligence. Diagnostics (Basel). 2025;15(5). 10.3390/diagnostics15050587. [DOI] [PMC free article] [PubMed]
- 31.Meo SA, Abukhalaf FA, ElToukhy RA, Sattar K. Exploring the role of DeepSeek-R1, ChatGPT-4, and Google Gemini in medical education: How valid and reliable are they? Pak J Med Sci. 2025;41(7):1887–92. 10.12669/pjms.41.7.12183. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Huh S. Are ChatGPT’s knowledge and interpretation ability comparable to those of medical students in Korea for taking a parasitology examination? a descriptive study. J Educ Eval Health Prof. 2023;20:1. 10.3352/jeehp.2023.20.1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Wang L, Mao Y, Wang L, Sun Y, Song J, Zhang Y. Suitability of GPT-4o as an evaluator of cardiopulmonary resuscitation skills examinations. Resuscitation. 2024;204. 10.1016/j.resuscitation.2024.110404. [DOI] [PubMed]
- 34.Ji Z, Lee N, Frieske R, Yu T, Su D, Xu Y, et al. Survey of Hallucination in Natural Language Generation. ACM Comput Surv. 2023;55(12):Article248. 10.1145/3571730. [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.