Healthcare AI Safety Gap · Don't break patients
    Peer-reviewed / official12 Nov 2026·US

    Primary-care patients rate a portal AI chatbot low on safety and trust

    Journal of General Internal Medicine: 2,678 primary-care portal users scored a proposed AI chatbot 2.51 out of 5 for patient safety and 2.59 for trust in accurate information, below its 3.30 for general acceptability.

    2.51 of 5
    mean patient rating of a portal AI chatbot's safety

    What the research found

    Researchers surveyed active primary-care patient-portal users at an academic health system about a proposed chatbot embedded in the portal, receiving 2,678 usable responses (16.1% response rate). On five-point scales patients gave general acceptability 3.30, trust in accurate health information 2.59, perceived patient safety 2.51 and willingness to talk to the tool 2.79. Technophobia and being retired were linked to lower ratings.

    Why it matters for providers

    Patients are telling providers the portal chatbot has to earn their trust on safety before it earns their use.

    Original source
    Journal of General Internal Medicine (academic health system patient-portal survey)
    Read the full report ↗
    Peer-reviewed or official source, the most reliable tier.