Generative AI · Education leadership

Actionable companion · Assessment and evidence of learning

From evidence to district action: Accuracy and Short-Term Consistency of Generative AI Chatbots in Guideline-Based Endodontic Decision Support: A Multi-Model Benchmarking Study

AI-assisted editorial implementation guide. The district actions are proposed applications, not findings or recommendations attributed to the original authors. No additional peer review is asserted.

Read original paperSearch paper agentCC BY 4.0

Evidence and context

Health-professions education context: test transfer to K–12 rather than assuming it.

Original authors’ abstract, reproduced under the linked source license:

Abstract Background/Objectives: Large language model (LLM) chatbots are increasingly used for dental information and decision support, yet their accuracy and short-term reproducibility in endodontics remain insufficiently established. This study compared five chatbots using open-ended questions derived from established AAE and ESE endodontic guidelines. Methods: Twenty-six guideline-based questions were content-validated by five endodontists using Lawshe’s Content Validity Ratio. ChatGPT-4o, Gemini 2.5 Pro, DeepSeek-V3-0324, ScholarGPT (academic version built on OpenAI’s GPT-4 architecture), and MedGebra GPT-4 answered each question across three days and three sessions per day, yielding 1170 responses. Two blinded endodontists scored responses on a 5-point guideline-concordance scale. Brunner–Langer LD-F2 analyses assessed model and temporal effects, while weighted kappa evaluated response consistency. Results: The overall model effect was significant (p < 0.001). ChatGPT-4o, Gemini 2.5 Pro, DeepSeek-V3-0324, and ScholarGPT showed comparable performance, whereas MedGebra GPT-4 performed significantly lower after Bonferroni correction. The day effect was not significant (p = 0.054), while the session effect (p = 0.048) and model × session interaction (p = 0.030) were significant. Weighted kappa values varied across models and assessment days, ranging from 0.689–0.730 for Gemini 2.5 Pro, 0.606–0.662 for ChatGPT-4o, 0.520–0.645 for DeepSeek-V3-0324, 0.458–0.592 for ScholarGPT, and 0.240–0.739 for MedGebra GPT-4. Conclusions: Guideline-aligned performance and short-term reproducibility differed across the evaluated chatbots, showing that accuracy and consistency represent distinct aspects of performance. Repeated assessment captured variation missed by single-session testing. These findings support guideline-based evaluation and clinician verification when generative AI chatbots are used to provide endodontic information relevant to decision support or dental education. Keywords: large language models, chatbots, generative artificial intelligence, endodontics, root canal therapy, decision support systems, clinical, practice guidelines as topic

A 90-day district pilot

  1. Days 1–15 · Appraise. Have the curriculum lead and a teacher read the methods and limitations. Record the study design, sample, comparator, outcomes and evidence gaps. Define a local question that the study can help inform.
  2. Days 16–30 · Design. Redesign one low-stakes assessment: require a first attempt, an AI-assisted revision, a source-check log and a short oral explanation. Use the same rubric for the comparison task. Name an accountable owner and set the baseline, success threshold and budget before implementation.
  3. Days 31–60 · Test. Start with a small voluntary cohort, preserve an accessible non-AI route and require teacher review. Use synthetic or de-identified examples; approve data handling and tool terms through district processes.
  4. Days 61–90 · Decide. Blind-score reasoning, factual accuracy and unaided transfer on a fresh task. Record suspected misconduct separately from demonstrated evidence. Compare with baseline, document uncertainty and stop or revise if learning worsens, access is unequal or critical errors persist.

The board decision brief

Report the learning problem, evidence quality, participating population, total cost including review time, measured benefit, errors, access differences and remaining uncertainties. Make a dated recommendation to stop, revise or expand. A positive perception survey alone does not establish learning improvement.

Attribution and reuse

Ilgın İlgenli; Ezgi Avcı; Timur Köse. Accuracy and Short-Term Consistency of Generative AI Chatbots in Guideline-Based Endodontic Decision Support: A Multi-Model Benchmarking Study. Dentistry Journal (2026-09-16). https://doi.org/10.3390/dj14090598

Companion prepared 2026-10-30. Original authorship and DOI are preserved. The abstract is an attributed source extract; the implementation plan is new AI-assisted editorial material, offered under the same CC BY 4.0 license. This guide has not undergone independent peer review.