Getty Images/iStockphoto

News

What patients want in AI-drafted patient portal messages

Patients are receptive to AI-drafted patient portal messages, so long as a clinician reviews them and they're informed that AI was involved.

Sara Heath
By
Published: 16 Jul 2026

Patients are warming to the notion of AI-drafted patient portal messages, with most saying in a small study in JAMA Network Open that AI could actually help them find answers to their questions sooner and save providers' time.

These findings come as industry experts work to fully understand the implications of embedding AI into patient portal messaging functions. This small study, which included qualitative interviews with 40 patients at an academic medical center, showed that AI could feasibly draft patient portal messages without damaging the patient-provider relationship.

Notably, the interviews demonstrated that patients consider portal messaging as transactional. Patients send secure direct messages to get an answer to a medical question, book an appointment or get a prescription refill -- not to deepen their relationship with their provider.

"Many patients recognized clinicians as 'busy' and interpreted AI-generated drafts as a means to help clinicians respond more quickly while managing competing priorities," the researchers wrote in the report. "In this framing, AI-supported messaging was viewed as mutually beneficial because it could reduce clinicians' burden while increasing the timeliness of patient responses."

But there are some key caveats.

For one thing, patients nearly universally insisted that AI-drafted patient portal messages be reviewed by a clinician before being sent. This reflects already accepted industry best practice.

Moreover, most patients want to be informed when AI has drafted a message from their provider. However, patients did not display a clear preference for how such disclosures are worded or when.

There's also the matter of tone, the qualitative data showed. When patients use the patient portal to perform a routine task, like refilling a prescription, they don't mind something that feels more clerical, as though written by AI.

"However, for messages involving uncertainty, anxiety, or potentially concerning test results, many participants expected clinicians to communicate with greater care and reassurance," the researchers explained. "For some participants, this meant that they would prefer clinicians to draft higher stakes messages or would prefer to use in-person communication methods."

Strategies for deploying AI-drafted patient portal messaging

Using these insights, the researchers developed a list of tips for implementing AI-drafted patient portal messaging.

Foremost, hospitals and health systems need to define key performance indicators and goals for AI-drafted messages, they suggested. They should also regularly evaluate these metrics and determine whether the technology is helping them hit their targets.

Next, organizations need to invest in mechanisms to ensure clinician review before hitting send. This must include a policy requiring human review, as well as EHR workflows that prevent clinicians from hitting send on messages that haven't yet been reviewed. Notably, organizations also need to establish a responsibility pathway for content included in AI-drafted patient portal messages.

In terms of how AI is deployed in patient portal messaging, the researchers recommended that organizations create contextual guidelines. This means offering advice on which clinical contexts AI should be used in and allowing for flexibility based on patient preference. Importantly, patient and family advisory councils should be involved in this process.

Finally, the researchers advised organizations to establish AI disclosure guidelines, including the messaging and where or when the disclosure occurs. Hospitals and health systems should periodically assess the disclosure guidelines for improvement, they said.

Do AI-drafted portal messages truly save clinicians time?

This advice hinges on the notion that using AI to draft patient portal messages actually saves clinician time. As observed during participant interviews, patients want speedy and informative answers to their messages, and they think AI could help make that happen.

This is important, as providers sift through more patient portal messages than ever. Earlier this year, a JAMA report showed a 153% surge in patient portal messages since 2020, adding to providers' already mounting workloads.

Some data has shown that AI can help support providers sort through their EHR inboxes. According to a 2025 report from NYU researchers, AI can reduce patient portal message response times by 7%.

However, that 7% only equates to about 20 seconds, the data showed.

That limited time savings is coupled with the fact that some experts think AI in patient portal messaging can actually add to providers' cognitive burdens. A recent report from Dartmouth found that AI rarely drafts responses the way a provider would, warranting extensive edits and potentially adding to workloads.

Of course, AI in healthcare is only in its infancy. As the industry continues to experiment with the technology, it will be critical to measure its genuine impact on the end-user experience.

Sara Heath is an executive editor at Xtelligent Healthcare Media, where she covers patient engagement, healthcare policy and health IT.

Dig Deeper on Patient data access