5 reasons patients check with a chatbot before—and often instead of—going to the doctor

Warning healthcare consumers about the unreliability of AI for medical advice is like screaming at tourists for frolicking in a decorative fountain on an oppressively hot and humid day. 

The person attempting the intervention sees nothing but the hazard and the heedlessness. He or she doesn’t grok the ignorability of the “No Bathing” signs vis à vis the depth of the demand for easy relief. 

Robert Shpiner, MD, of UCLA Health perceives a similar dynamic affecting internal medicine practitioners—and presumably all primary care providers—who try to scare off patients from turning to AI chatbots first and the doctor only after that, if at all. 

Of course, in 2026, that’s a lot of patients. 

The shift is easy to blame on the technology itself, Shpiner suggests, but it is also about internal medicine.

“The same conditions sending patients to consumer AI tools,” he writes, “have made ordinary care harder to access, navigate and trust.”

It’s a brief yet thought-provoking opinion piece. Annals of Internal Medicine published it online July 27. 

Here are five takeaways. 

1. Chatbot use reflects both health system strain and a broader shift in how patients seek information.

‘The patient who turns to AI for convenience also forms a habit that competes with the clinical relationship.’

2. The attraction is not that the tool is wise but that it is available.

‘Often the patient is responding to the lived reality of modern care: no timely appointment, no established clinician, no confidence that a portal message will be answered and no wish to generate an urgent care bill for a question that may be minor.’ 

3. Policy should reach beyond model governance, building on what exists.

‘Even excellent governance will not solve the problem if patients use chatbots to compensate for failures in access, affordability, continuity and respect.’

4. Trust is less a sentiment than an operational property of care.

‘It erodes when every question is routed through hold music, billing opacity, prior authorization or a portal that functions as a dead-letter box. No campaign about chatbot risk will outweigh those experiences.’

5. Patients should still be warned that general-purpose chatbots are often unreliable for medical advice—true and necessary. 

‘Yet the larger lesson is harder because it is not mainly about the software.’ 

“Patients ask the chatbot first because, for too many, it feels easier to approach than the healthcare system,” Shpiner concludes. “Until internal medicine treats this as a problem of structure and trust rather than only of misinformation, the warnings will continue—and so will the behavior.”