Actionable companion · Assessment & reasoning

From AI evidence to responsible practice: Comparing ChatGPT-3.5, Gemini 2.0, and DeepSeek V3 for pediatric pneumonia learning in medical students

AI-assisted editorial implementation guide, separate from the original research. Proposed actions are not findings or recommendations attributed to the source authors. No independent peer review of this guide is asserted.

Read original paper Search paper agent CC BY 4.0

Evidence and context

Title and abstract screening; stage labels describe reported settings, not verified numerical ages. Design labels require full-text appraisal. Reported stage(s): Higher education. Health-professions education findings do not establish effectiveness in K–12 classrooms.

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

Pediatric pneumonia (PP) remains an important topic in undergraduate medical education and offers a suitable framework for evaluating large language models (LLMs) in AI-assisted learning. We developed a 27 open-ended survey including five core domains of PP, such as diagnosis, etiology, diagnostics, treatment, and prevention. DeepSeek V3, Gemini 2.0, and ChatGPT-3.5, were each provided with identical reference materials. Two pediatric infectious disease specialists independently assessed responses using a structured 10-point rubric through Licert, which presents a custom evaluation tool. DeepSeek V3 achieved the highest mean score (9.9), outperforming ChatGPT-3.5 (7.7), and Gemini 2.0 (7.5) through all domains (p < 0.001). Moreover, it received a full score in 26 out of 27 questions (96.3%) and achieved an accuracy score of ≥ 5. In addition, the highest performance appeared in higher-order reasoning areas, including age-specific etiology and imaging interpretation, with DeepSeek V3 outperforming others by up to 3.2 points. While all models demonstrated almost safety, the variability in content quality however highlights the necessity for careful platform selection. Therefore, future research should explore educational outcomes comparing AI-assisted and conventional learning approaches to better define the role of LLMs in medical education. Supplementary Information The online version contains supplementary material available at 10.1038/s41598-025-27722-2.

What kind of evidence is this?

Supporting AI learning science · Survey / observational study

Stage: Higher education. Screening labels describe reported context and study-design language. They are not a risk-of-bias rating or proof of effectiveness for an age band.

Study methods · source passage

This is a cross-sectional and comparative evaluation study designed to investigate the performance of three LLMs, including ChatGPT-3.5 (OpenAI), Gemini 2.0 (Google), and DeepSeek V3 in providing educational content focusing on PP. The primary objective is to determine the accuracy, completeness, and safety of these models to questions, which are commonly used by medical students during their medical training. To simulate a controlled learning environment, we provided each model with the full text of the Community-Acquired Pneumonia chapter from ‘Nelson Textbook of Pediatrics’11, the ‘Revised WHO classification and treatment of childhood pneumonia at health facilities publication’12, and the 2011 IDSA guideline13 for the management of community-acquired pneumonia in children. These resources represent as the most relevant reference materials available to the models. This study did not involve any human participants, including medical students. All analyses were performed using model-generated outputs. An ethics or institutional review board approval was thus not necessary.

Limitations and uncertainty · source passage

Pediatric pneumonia (PP) remains an important topic in undergraduate medical education and offers a suitable framework for evaluating large language models (LLMs) in AI-assisted learning. We developed a 27 open-ended survey including five core domains of PP, such as diagnosis, etiology, diagnostics, treatment, and prevention. DeepSeek V3, Gemini 2.0, and ChatGPT-3.5, were each provided with identical reference materials. Two pediatric infectious disease specialists independently assessed responses using a structured 10-point rubric through Licert, which presents a custom evaluation tool. DeepSeek V3 achieved the highest mean score (9.9), outperforming ChatGPT-3.5 (7.7), and Gemini 2.0 (7.5) through all domains (p < 0.001). Moreover, it received a full score in 26 out of 27 questions (96.3%) and achieved an accuracy score of ≥ 5. In addition, the highest performance appeared in higher-order reasoning areas, including age-specific etiology and imaging interpretation, with DeepSeek V3 outperforming others by up to 3.2 points. While all models demonstrated almost safety, the variability in content quality however highlights the necessity for careful platform selection. Therefore, future research should explore educational outcomes comparing AI-assisted and conventional learning approaches to better define the role of LLMs in medical education. Supplementary Information The online version contains supplementary material available at 10.1038/s41598-025-27722-2.

Design the system before the pilot

Use this study to inform a learning mechanism or implementation question. It is not direct evidence that autonomous agents improve learning. Keep the proposed agent behavior separate from what the source actually tested.

Use common rubrics, blind scoring where feasible and a fresh unaided reasoning task. Document permissible assistance and avoid treating AI-detector scores as proof of misconduct.

A 90-day district pilot

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.
Days 16–30 · Design. Choose one disciplinary task. Require a first attempt, a documented AI consultation, a critique of the output and an independent explanation. Align the comparison group and scoring rubric with the same objective. Name an accountable owner and set the baseline, success threshold and budget before implementation.
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.
Days 61–90 · Decide. Blind-score reasoning and accuracy on a new task without AI. Distinguish model examination scores, student perceptions and actual learner gains; only the last directly tests educational benefit. 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

Kubra Aykac; Osman Cubuk; Osman Oguz Demir; Young June Choe; Malik Aydin; Yasemin Ozsurekci. Comparing ChatGPT-3.5, Gemini 2.0, and DeepSeek V3 for pediatric pneumonia learning in medical students. Scientific Reports (2025-11-18). https://doi.org/10.1038/s41598-025-27722-2

Companion prepared 2026-11-01. 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.