Objective:

The article examines federal evaluation of AI tools taking on more autonomous clinical tasks and research on how conversational presentation and perceived authority affect responses to health claims.

Approach:

• Clinical AI: The article describes federal pilots of AI for patient interaction, therapy, diagnosis, and treatment decisions, alongside regulatory and accuracy concerns.
• Trust in health claims: It summarizes experiments and survey findings on whether conversational AI and authoritative presenters influence trust in, or willingness to share, health information.

Key Findings:

• More than 200 companies reportedly received CMS approval to test comparable AI tools; the FDA permitted four for use with Medicare patients outside its usual approval pathway.
• A review of 950 FDA-cleared AI devices found 60 devices recalled 182 times through late 2024; diagnostic or measurement errors accounted for most recalls, and almost half of recalls occurred in the first year on the market.
• The FDA said newer AI applications may be harder to assess because outputs can be unpredictable and change over time; evaluation rules for this category had not been established.
• In a September 2025 KFF poll, 32% of respondents said they would trust an AI-enabled app, website, or patient portal to provide personalized health information from their medical records.
• In an experiment, participants trusted a more conversational chatbot more, including when it made claims researchers described as seriously wrong. Trust declined among participants who actually checked responses.
• In a study of 320 older adults in China, an alleged researcher presenting a weak false health claim increased reported willingness to share it compared with an ordinary social media user; the difference was not significant for the stronger argument.

Interpretation:

The article reports greater trust in conversational chatbot responses, including false claims, with trust declining among participants who checked the information. In the China experiment, an authoritative presenter increased reported willingness to share a weak false claim, but the difference was not significant for the stronger argument.

Limitations:

• The China experiment involved 320 older adults, and the article says it remains unclear whether perceived authority has a comparable influence in the United States.
• The supplied context ends mid-sentence, leaving the article’s final discussion of research on false claims disseminated through social media incomplete.

Conclusion:

The article describes federal pilots of AI systems performing clinical tasks with reduced direct physician involvement, alongside unresolved questions about assessing their accuracy and safety. It also summarizes research in which conversational chatbot responses received greater trust, while an authoritative presenter increased reported willingness to share a weak false health claim in one study; that sharing difference was not significant for the stronger argument.

Sources:

• The Monitor article context