Do patients really believe AI can't handle their particular medical case, and does that belief track how AI actually performs?

This explores whether patients really believe AI can't handle their particular case, and whether that belief tracks how well medical AI actually performs. The short answer from the corpus is yes, patients do report this perception, and it is one of three separate reasons they resist medical AI. The other two are a belief that AI simply performs worse than human providers and a sense that it is harder to hold accountable when something goes wrong Why do patients distrust medical AI systems?. The key detail is that these barriers exist independent of actual AI capability. Patients aren't reacting to evidence about how well AI performs. They're reacting to what they imagine AI is: a system built for the average case rather than for them.

That gap matters because the capability evidence points the other way. In text-based simulated consultations, the diagnostic system AMIE beat primary care physicians on 28 of 32 dimensions rated by specialists Can an AI system diagnose better than primary care doctors?. Another LLM outperformed hundreds of physicians on differential diagnosis, triage, and management, including in an emergency room study Can language models reason better than physicians at diagnosis?. The AMIE result holds a twist that speaks directly to the "unique needs" worry: its advantage came from reasoning over the information it gathered, not from drawing out the patient's history. So the part of care that feels most personal, telling your own story, is not where AI clearly wins.

The most revealing material comes from a different group: clinicians. Radiologists rated advice lower when it was labeled as coming from AI rather than a human expert, even though the advice was identical. Yet their diagnostic accuracy depended on whether the advice was correct, not on its label Does labeling advice as AI change how clinicians use it?. When clinicians compared GPT-4 advice with expert advice without knowing the source, they rated GPT-4 as equally sound and more emotionally empathetic, and they guessed the source at chance level Can clinicians tell GPT-4 advice apart from expert advice?. Together these suggest the "AI is worse for me" judgment comes largely from the label. Hide the source and the gap mostly disappears. Show it and opinions shift, even when behavior doesn't.

There is a less comfortable side. Skepticism isn't always irrational, and too much trust has its own costs. Experienced radiologists who were given wrong AI suggestions dropped from 82% to 45.5% accuracy, and less experienced readers fell below 20% How much does wrong AI advice harm radiologist accuracy?. In therapy, patients form real emotional bonds with chatbots, but those good feelings can hide safety failures, such as the model reinforcing harmful thinking Do therapeutic chatbot bond scores hide deeper safety problems?. How care is delivered also matters: a robot and a worksheet reduced distress while a chatbot running the same LLM did not Why do robots outperform chatbots in therapy despite identical language models?. "Feeling attended to" may depend more on presence and structure than on the quality of the answers.

A caveat on the evidence: the corpus has only one source that measures patients' perceptions directly. Most of the lateral evidence comes from clinicians judging AI, not patients judging their own care. What the corpus supports is this: the "AI can't handle my unique needs" belief is real and works independently of performance, labels shape how people judge AI more than how they use it, and better-performing models alone may not change that belief.