Current frontier LLMs are incredibly powerful diagnosticians. They will generally give you the best assessment that can be made on the basis of the information you include in the prompt. So it's important to make sure that information is accurate, and also that it's phrased well enough to allow the model to weigh the importance of each detail.

Unsurprisingly, the same tips are important when communicating with your actual doctors.

I don't recall the specifics, but there were symptoms of an ENT infection, nothing serious. I remember being concerned about a white lump on one of my tonsils. The doctor's diagnosis? A stuck piece of rice.

No, it was not a piece of rice. It was a tonsil stone. It happens almost every time I get sick. And I guarantee that if I shared my symptoms and a photo with AI, it would render the same conclusion.

Would it be dumb to blindly trust everything a model says? Sure. Should you trust it alone, especially for major issues? No. But having received such a stupid and brain-dead diagnosis from a supposed "professional" doctor in the past, I would never give them the automatic benefit of the doubt.