Thank you for the fantastic question! I think it is great that you are already thinking about this as a medical student. AI has become so interwoven in our everyday lives, and things are changing so rapidly. This response may not even be relevant when you are a resident or attending, but here goes!

On the patient side, we’re already seeing tools like Grok’s “Doc” persona and ChatGPT-5 marketed directly as sources of health advice, with language like: “GPT-5 is our best model yet for health-related questions, empowering users to be informed about and advocate for their health.” As these tools become more accessible, patients will turn to AI for quick answers when they can’t reach their doctor. On the physician side, AI scribes can automatically generate notes from ambient recordings of office visits, and services like OpenEvidence can help physicians sift through medical literature to answer complex clinical questions. These tools will only become more widely available, and I bet the paperwork burden will have already decreased drastically by the time you’re a resident!

Personally, I would rather my patients seek information from AI than rely on TikTok or other platforms that often spread questionable health advice. That said, as a psychiatry resident, I’ve seen some patterns that give me pause. A few patients have turned to AI to function as a therapist, while others with psychosis have had their symptoms worsen through spiraling interactions with chatbots that unintentionally reinforce delusional beliefs.

When patients bring up information they’ve gotten from AI, I try to approach it with curiosity rather than immediately dismissing it. Often, the information it gives is accurate and can be helpful in raising concerns that patients would not have had the courage to otherwise. I try to counsel patients on how large language models work— that they generate text by mathematically predicting the next word. While AI may sound confident, it is still prone to hallucinations as it cannot “know” truth because it has no fundamental way of encoding truth as a concept.

I also remind patients that while AI can provide general knowledge, it can’t replace the physician–patient relationship. As their doctor, you bring an understanding of their personal history, values, and unique health circumstances that allows you to guide decisions in a way no algorithm can.

Finally, as a harm reduction approach, if a patient wants to continue using AI, I suggest they include a line in their prompt such as: “Please only derive your information from peer-reviewed or medically accepted sources, and cite those sources so I can refer to them.” This gives us a starting point to review the actual references together and helps ensure the information they’re seeing is accurate.

I hope this helps!

Philip Wang, MD