Guides
When asking ChatGPT becomes part of the loop.
You ask, it answers, and for about ten minutes the question is settled. Then a new wording of it turns up, and you are back. If that is familiar, this page is about what is actually happening, and what helps.
The pattern, named.
It usually is not one question. It is the same question reworded, again and again, until the answer finally sounds right, and then once more later to check it still does. Some people describe confessing things to the chat to get them off their chest. Many describe doing it at 3am, because that is the hour the doubt is loudest and a chatbot is the only thing awake.
Asking for certainty like this has a clinical name: reassurance-seeking, and for people with obsessive-compulsive disorder it works like any other compulsion. The relief is real, and it is rented rather than ownedNOCD. Googling the worry, replaying the memory and asking someone you love are the same move through different doors. The chat window is just the newest door.
Why it only holds for ten minutes.
Every time the doubt spikes, there are two ways down. Get the answer and the feeling drops fast, but the lesson your brain takes is that the question was dangerous and answering it is what made you safe. So it brings the question back, and it brings it back louder NICE CG31 (full).
That is why the certainty never lasts. The problem was never a missing fact, so no fact settles it. The next what-if only has to be worded slightly differently to get past the answer you already have.
A chatbot is the most patient answer machine ever built.
A person eventually gets tired of the question. A chatbot never does. It answers instantly, at any hour, takes every rewording seriously, and goes into the specifics in a way a search result cannot. Everything that makes it a good assistant makes it a perfect reassurance dispenser.
This is not a private worry. A 2026 study read a hundred posts from the main OCD community on Reddit about chatbot use and found the same story told over and over: reassurance on tap, and the loop growing around it. OCD clinicians have written the same warning.
One more thing people realise late: those conversations are not private the way they feel at 3am. They sit readable on a server, tied to an account.
What helps instead.
The exhausting part is the fight: arguing with the thought all day, patrolling for the next what-if, holding the case files open. What published clinical guidance recommends first is practice in the other direction, a way of working called exposure and response prevention NICE CG31IOCDF. In plain words: let the question stand unanswered, on purpose, and stay while the feeling peaks and comes down on its own. Each time you do, the thought loses a little of its authority. We have a from-scratch explainer on how that works.
Small versions count. Noticing the urge to reword the question and waiting ten minutes before asking is practice. So is asking once and refusing the follow-up. The aim is never to win the argument with the thought. It is to stop attending the argument, and to get on with your day while the maybe stands.
A therapist trained in this is the best thing there is. If you can get to one, go; the resources page lists official places to start, free to everyone.
Maybe is built for exactly this loop.
It is short lessons and an AI chat. The lessons show you the ropes: why thoughts stick, why answering them keeps them alive, and how facing things on purpose loosens the grip. The chat backs you up when it hits, and it will not answer the question, however many ways you word it. Everything you write is sealed on your device before it is sent, so nobody can read it, including us.
Seven days free, then $19 a month. Card up front, first charge on day seven, cancel any time before then. If you do pay and it isn’t for you, 30 days to get it back.
Structured practice, not therapy and not a crisis service. If tonight is an emergency, the safety page lists people who answer the phone.