AI-generated virtual patients exhibit limitations including inauthentic conversational flow, excessive agreeableness, role departure, and clinical inaccuracy¶

Claim · All claims
Evidence · 1 study (2 entries) · review r2–r3 · q2–q3
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person · recheck by 2029: tool and model accuracy changes quickly

Subclaims¶

q3 i? Reported limitations include lack of natural conversational flow, lack of emotional depth, excessive agreeableness, overly polished answers, deficient clinical accuracy, and inappropriately departing from the intended patient role, such as prematurely giving away the diagnosis. → Fang Eric Kin Meng 2026 q3 i? Pre-2023 rule-based systems showed limitations centered on unrealistic dialog with limited response ranges, with role departure less prominent; multimodal additions introduced latency, limited nonverbal range, and speech-to-text inaccuracy. → Fang Eric Kin Meng 2026 (2)

Evidence¶

Fang Eric Kin Meng 2026¶

Fang Eric Kin Meng, Boon Yuru, Ho Jin Yang, Tey Natasha, Low Matthew Jian Wen. (2026). Artificial intelligence generated virtual patients in health professions education: a scoping review. BMC Medical Education. https://doi.org/10.1186/s12909-026-10028-w

q3 · i? · review · r3

Qualitative synthesis of limitations charted across 97 included studies of AI-generated virtual patients in HPE. The review reports limitations including "lack of natural conversational flow", "being excessively agreeable", and role departure such as prematurely giving away the diagnosis like a tutor.

"Limitations included a lack of natural conversational flow [48], a lack of emotional depth [96, 125], being excessively agreeable [10, 37, 98], giving overly polished and complete answers"

Fang Eric Kin Meng 2026 (2)¶

Fang Eric Kin Meng, Boon Yuru, Ho Jin Yang, Tey Natasha, Low Matthew Jian Wen. (2026). Artificial intelligence generated virtual patients in health professions education: a scoping review. BMC Medical Education. https://doi.org/10.1186/s12909-026-10028-w

q2 · i? · review · r2

Subgroup analysis of studies published before 2023, prior to the LLM era: the review reports limitations "centered around unrealistic dialog" from limited response ranges in rule-based scripted systems, with role departure a less prominent concern; multimodal systems added latency and input-recognition challenges.

"prior to the LLM era, limitations centered around unrealistic dialog, especially due to virtual patients having a limited range of responses and related difficulty in adapting to the context of the conversation"

Discussion¶

Related Claims¶

• ChatGPT-generated text shows inaccuracy, weak logical flow, lack of critical elaboration and non-originality in an author's first-hand test — related
• Students report negative experiences with ChatGPT including inaccurate outputs, lack of emotional connection, and risk of overreliance, which constrain competence — related
• AI use also carries negative outcomes: technical instability, negative emotional impacts, over-reliance, and contextual contingency — related
• Persistent challenges include feedback inaccuracy, learner overreliance, and limited transparency — related
• Learner-level challenges of ChatGPT in STEAM IBL include over-reliance, inaccurate or irrelevant feedback, conceptual confusion, and diminished creativity — related
• AWE feedback has limited depth and questionable accuracy for evaluating complex writing aspects — related
• AI-generated virtual patients in HPE are used mainly for teaching and learning communication skills, diagnostic reasoning and management reasoning, with assessment a minority use case — related
• Outcome evaluation of AI-generated virtual patients in HPE is concentrated at Kirkpatrick level 1, with no studies reporting level 3 or 4 outcomes — related
• Reviewed GenAI virtual patients show underdeveloped emotional-behavioral complexity, limited adaptability beyond dialog, and no memory across interactions — a narrower finding that bears on this claim