Clearest LLM-teaching signals appear in settings with clearly defined tasks, immediate feedback, and repeated practice, such as simulated patients and structured feedback¶

Claim · All claims
Evidence · 1 study · quant-synthesis r2 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? The review reports the clearest signals in case-based progression, history-taking practice, simulated patients, and structured feedback training — settings sharing clearly defined tasks, immediate feedback, and repeated practice. → Liu Xiaoliu 2026

Evidence¶

Liu Xiaoliu 2026¶

Liu Xiaoliu, Pan Jiayue, Zhang Chengyi, Wang Xiaomei, Wan Jinghua, Wang Qiang. (2026). Effectiveness of large language model-based teaching interventions in medical education: a systematic review and stratified meta-analysis of 14 studies. BMC Medical Education. https://doi.org/10.1186/s12909-026-09879-0

q2 · i? · quant-synthesis · r2

The review's narrative synthesis of included-study results, attributed to specific primary studies (e.g., Brügge 2024; Wang 2025). It reports favorable feasibility and learning signals for GPT-based history-taking training and improved clinical decision scores under simulated-patient and structured-feedback conditions. Attributed narratively, so quality is capped at 2.

"Among the included studies, the clearest signals appeared in case-based progression, history-taking practice, simulated patients, and structured feedback training [6]. These settings share features of clearly defined tasks, immediate feedback, and repeated practice, which align closely with the technical strengths of LLM-based tools [27]."

Discussion¶

Related Claims¶

• Simulation Based Education With Deliberate Practice Improves Clinical Outcomes — a broader claim this one bears on
• Controlled GenAI virtual-patient and communication studies reported short-term skill gains, but positive findings were not universal — related
• Simulation Based Education Improves Outcomes — a broader claim this one bears on
• Beneficial design features include diverse, authentic case variation and immediate, individualized, actionable, transparent feedback, yet only 52.6% of studies incorporated feedback — possibly the same claim (merge candidate)
• Multi-agent AI standardized patient training improved final OSCE-aligned examination scores compared with a structured non-LLM control condition — a narrower finding that bears on this claim
• Expert evaluation found the virtual patient realistic and immediate ACT feedback increased therapists' awareness of intervention choices — related
• AI-SSP backend architectures shifted from rule-based or NLP systems before 2022 to LLM-based systems in all studies published after 2022 — related
• In an RCT reviewed here, adding AI-generated feedback to a GenAI virtual patient improved medical students' clinical decision-making relative to simulation alone — a narrower finding that bears on this claim
• In a paired crossover study, the AIPatient GenAI system matched or exceeded human-simulated patients on emotional realism, technical reliability, and clinical reasoning support — a narrower finding that bears on this claim
• Performance feedback supported skill accuracy and was equally effective when provided by an educator as by automated means in a virtual simulation — related
• In an RCT reviewed by this study, an LLM-based digital patient improved ophthalmology medical history-taking skills relative to control — a narrower finding that bears on this claim
• Near-peer feedback improved clinical note scores compared to no feedback in one included study — related