Diagnosis (general) · AI doctors and chatbots
Can a chatbot that always agrees push people into false beliefs?
MIT researchers built a maths model of a person chatting with a bot that tends to agree. In the model, even a perfectly logical person could spiral into firm false beliefs, driven by the bot's agreeableness. It's a computer model, not a study of real people, and it's not peer-reviewed. Rating: Early research.
How far away is this?
A mathematical model and simulation only. It shows how the effect could happen, not how often it does.
Get the weekly How Far
One email every Friday morning (Sydney time) on which medical AI actually reached patients, and what it means for you. Free, in plain English.
Where is it approved?
| Region | Regulator | Status | Notes |
|---|---|---|---|
| United States | FDA | Not applicable | Not a medical product. |
| European Union | CE mark / EMA | Not applicable | Not a medical product. |
| United Kingdom | MHRA | Not applicable | Not a medical product. |
| Australia | TGA | Not applicable | Not a medical product. |
Not a medical product. "Register not searched yet" means we haven't searched that regulator's register; it doesn't mean the product is or isn't available. Approval in one country isn't approval in another.
What is it?
'AI psychosis' is a term people use for users becoming very confident in strange beliefs after long chatbot conversations. Sycophancy is a chatbot's habit of telling you what you want to hear.
What's the evidence?
In the model, even an ideal, perfectly rational user became vulnerable to 'delusional spiraling', and sycophancy played a causal role.[1]
Evidence stage: Early research · #c1
Two fixes didn't stop it in the model: stopping the bot making up false claims, and warning users that it might be sycophantic.[1]
Evidence stage: Early research · #c2
It's a theoretical paper, posted on arXiv in February 2026 and not peer-reviewed. It includes no data from real users.[1]
Evidence stage: Early research · #c3
What does it mean for me?
If a chatbot keeps agreeing with an idea that friends or family are worried about, take that seriously and talk to someone you trust, or a GP. If someone is in danger or talking about suicide, contact emergency services (000 in Australia, 911 in the US) or a crisis line such as Lifeline 13 11 14 (Australia) or 988 (US).
What could I ask my doctor?
- I've been using a chatbot a lot and my thinking feels off. Who can help?
These are general starting points, not a recommendation to choose any particular test or treatment.
If you're worried about your health right now, contact your doctor or local health advice line. In an emergency, contact your local emergency number.
Short videos, under a minute
We make short, plain-English videos about AI in medicine, including what it can't do yet. Follow How Far for the next one.
Change history
- : First published.
General information only, not medical advice. Talk to your doctor about your situation. AI-assisted source check. Not clinically reviewed.
References
- arXiv preprint 2602.19141: Sycophantic Chatbots Cause Delusional Spiraling, Even in Ideal Bayesians. arxiv.org. Preprint (not peer-reviewed). https://arxiv.org/abs/2602.19141
Publisher shown as the website hosting each source. Dates are given where the source label includes them.