JMIR Publications

ABSTRACT

Background:

Generative artificial intelligence (GenAI) tools powered by large language models (LLMs) are increasingly used by the public to seek health information. Unlike traditional web search, GenAI systems generate conversational answers, which may influence how users assess credibility, manage uncertainty, and decide whether to verify information or consult clinicians. Evidence is needed to clarify facilitators, barriers, and user practices in GenAI-supported health information seeking.

Objective:

This systematic review synthesizes empirical research on consumer and patient health information seeking with GenAI/LLM tools, focusing on study contexts, adoption and use outcomes, facilitators and barriers, with implications for clinician-patient interactions.

Methods:

We conducted a review following PRISMA-ScR. Records were identified through database searching and screened using predefined eligibility criteria. Included studies were extracted using a structured form capturing study characteristics, GenAI tool type, health context, outcomes, as well as facilitators and barriers. Findings were synthesized using structured grouping aligned to the RQs.

Results:

The review included 27 studies. GenAI was used for symptom appraisal, condition understanding, treatment options, and care navigation. Facilitators emphasized convenience and clarity, including efficiency and access (29.6%, n=8), comprehensibility and presentation quality (40.7%, n=11), personalization and specificity (18.5%, n=5), and affective or interpersonal comfort (18.5%, n=5). Barriers were dominated by credibility and trust concerns (48.1%, n=13), particularly when accuracy cues or citations were missing or difficult to interpret. Additional barriers included perceived unsuitability for complex, urgent, or emotionally charged situations (18.5%, n=5), privacy or data security concerns (14.8%, n=4), limited prompting skills (7.4%, n=2), and modality or interaction constraints that hindered credibility assessment and information comparison (18.5%, n=5). Literacy-related capability was reported in 22.2% of studies (n=6), and verification-supporting features (e.g., visible sourcing, transcripts, and save/revisit/share functions) were reported in 18.5% of studies (n=5).

Conclusions:

GenAI is used for diverse health information needs, but reliance is shaped by trust, perceived risk, and verification capacity. Future research should improve reporting of tools and prompting conditions, standardize measures of reliance and verification, and evaluate use in higher-stakes and underserved contexts to inform safer design and public guidance.