Asking ChatGPT for reassurance: why the relief never lasts

Asking ChatGPT for reassurance about whether you’re a bad person, have OCD or are safe can bring a few minutes of relief before doubt returns. A chatbot is always available and will answer the same question a hundred different ways, which can keep reassurance seeking going all day. What helps is noticing and delaying the urge, practising uncertainty and, if the habit is part of OCD, getting help through exposure and response prevention.

Why does the chatbot’s answer stop working so quickly?

The answer stops working because the question was never really about information. When a doubt like “Did I hurt someone?” or “Do I secretly want this thought?” sticks, what drives it is the feeling of not being completely sure. A reply can deal with the words of the question, but it can’t switch off that feeling, so the relief fades and the doubt finds a new angle.

The relief itself is part of the problem. Each time asking makes you feel better, even for a minute, your mind learns that asking is how to cope with doubt. The next doubt arrives sooner, the urge to ask is stronger, and the calm lasts a little less long. Psychologists who work with OCD and anxiety have described this pattern for decades, long before chatbots, with people asking partners, parents, doctors and search engines.

A chatbot adds a few things that make the loop easier to fall into:

• It’s always available. At three in the morning, in a work meeting, on the bus. There’s never a pause long enough for the urge to fade on its own.
• There’s no social cost. Asking a friend the same thing for the tenth time feels embarrassing, which acts as a natural brake. A chatbot never sighs.
• It answers endlessly and politely. You can ask the same question in twenty different ways and get twenty careful replies.
• It hedges. A responsible answer often includes “I can’t say for certain” or “only a professional can tell you”. For a mind looking for certainty, that one line can undo everything else in the reply.
• It doesn’t know you. Even a perfect answer leaves room for “but it doesn’t know the whole story”, which sends you back to type the whole story in.

Is asking AI for reassurance a compulsion?

It can be. In OCD, a compulsion is anything you feel driven to do to bring down distress or to stop something bad from happening, done again and again even though part of you knows it doesn’t settle things for long. Asking for reassurance is one of the most common compulsions, alongside checking, washing and mental rituals, and asking a chatbot is simply a new place to do it.

Not every question you type is a compulsion, though. What matters is why you’re asking and what happens after.

If the right-hand column sounds familiar, and the doubts are about harm, contamination, morality, your relationship or whether things are “just right”, they may be part of an OCD pattern. The unwanted thoughts and rituals check on Sprout looks at four common kinds separately, including asking for reassurance, and shows which ones take up most of your time. It isn’t a diagnosis, but it can show whether talking to someone is worth it. Taking the test is free; results need Sprout Plus (7 days free).

What does a reassurance loop with a chatbot look like?

A chatbot reassurance loop means repeatedly asking about the same doubt without feeling settled for long. Common signs include:

• Rewording the same question until the answer feels right, then asking once more to make sure.
• Opening a new chat so it “doesn’t know” you’ve already asked, or asking several different apps and comparing their answers.
• Typing out long, detailed descriptions of a thought or an event to get a verdict on it.
• Asking it to promise: “Just tell me I’m not a bad person.”
• Scrolling back through old chats to reread a reassuring reply.
• Feeling worse, not better, when the answer includes any doubt at all.
• Losing an hour, or an evening, to one conversation.
• Keeping it hidden because you know how many times you’ve asked.

The topics vary. Some people ask whether an intrusive thought means they’re dangerous. Some ask whether they really love their partner, whether a symptom is serious, whether they did something wrong at work, or whether a past event “counts”. The content changes; the loop doesn’t.

The confession version

Some people don’t ask a question at all. They type out a thought or a memory in full, as a kind of confession, and wait for the reply to tell them it’s okay. It can feel like getting something off your chest. If you need to do it again the next day, about the same thing, it’s working as a ritual rather than a release. For why some thoughts get stuck like this, see unwanted, intrusive thoughts.

Why can’t any answer give you certainty?

Because the kind of certainty the doubt asks for doesn’t exist. Nobody, human or machine, can promise with complete confidence that you will never act on a thought, that you locked the door, or that you’re a good person. A careful chatbot says as much, and a careless one might sound sure and be wrong, which is worse.

Chatbots also aren’t consistent. Ask the same thing twice and you may get two slightly different answers. For most people that doesn’t matter. For someone checking for certainty, a small difference between two replies becomes a new thing to resolve.

Chatbots are new, and there isn’t yet an established body of research on reassurance seeking from them in particular. What is well established is how reassurance seeking works in OCD and anxiety in general, and there’s no obvious reason a chatbot would be the exception.

What can you do instead when the urge to ask hits?

Delay asking and leave the doubt unanswered for a while. The delay can feel uncomfortable, but it teaches your mind that the discomfort fades on its own.

Name the urge

“This is the reassurance urge” is enough. You’re not deciding whether the doubt is true, only noticing what kind of thing it is.

Set a timer before you ask

Pick a delay you can manage, even five minutes, and do something else while it runs. If the urge is still there when it ends, try another five. Many people find the urge has dropped by the time the second timer goes off.

Answer with “maybe”

Instead of arguing the doubt away, try agreeing that it can’t be settled: “Maybe, maybe not. I’m going to carry on with my evening anyway.” This is practice at living with uncertainty.

Cut down rather than stop overnight

If stopping completely feels impossible, set limits. One question a day, no rewording, no second app, no rereading old chats. Each limit you keep is progress.

Make it a little harder to ask

Move the app off your home screen, log out at night, or leave your phone in another room for the first hour of the evening. Small barriers give the urge time to pass.

Be kind to yourself when you slip

You will ask again sometimes. Beating yourself up about it adds another layer of distress, which makes the next urge stronger. Notice it, and start the next delay. The three-minute self-compassion pause can help when the guilt is loud.

The same approach works for reassurance from people. If you often ask a partner or parent, it can help to tell them you’re cutting down, and agree together how they’ll respond: kindly, but without answering the question. When the doubts are mostly about whether a partner loves you, rather than about your own thoughts, anxious attachment patterns help explain it.

How does exposure and response prevention deal with reassurance?

Exposure and response prevention (ERP), a form of cognitive behavioural therapy, is the best-established psychological treatment for OCD. With a therapist, you gradually face the thoughts and situations that set off the doubt, and practise not doing the compulsion, including not asking for reassurance, at a pace you agree on together.

For reassurance, that might mean writing down a feared thought and leaving it unanswered, letting a doubt sit for an hour, then a day, or saying “maybe” out loud to a question you’d normally need settled. It starts with things that feel manageable and builds up. Over time, the brain learns that the anxiety rises and falls without the answer, and the urge to ask loses its grip.

A therapist also brings something a chatbot can’t: a plan that holds across weeks, someone who recognises the same doubt behind a new question, and support when it gets hard. A doctor can talk you through treatment options and how to find ERP where you live.

Can AI be useful at all if you have OCD or anxiety?

Yes, for learning rather than for certainty. A chatbot can explain what ERP is, help you write down what you want to tell your doctor, suggest questions to ask a therapist, or help you practise a hard conversation. Those are things you read once and use.

Where it goes wrong is when you ask it for a verdict on yourself: whether you’re dangerous, whether you did something, whether a thought means anything, or whether you have a condition. A simple test: if you’d be happy to read the answer once and close the app, you’re learning. If you’ll need it again tomorrow, it has become reassurance.

Two practical points. A chatbot can’t diagnose you and isn’t a replacement for treatment. And what you type is usually stored by the company that runs it, so it’s worth checking its privacy settings before pasting in your most private thoughts.

When should you seek help for reassurance seeking?

It’s worth talking to a doctor or therapist when the asking takes up a lot of your day, when it leaves you more anxious overall, or when the doubts are pulling you into other rituals or making you avoid people and places. You don’t need to know whether it’s OCD first. Describing the thoughts and what you do about them is enough to start.

A short questionnaire can help you put it into words. The OCD questionnaire takes about seven minutes and covers the last month. If your worries are less about shocking thoughts and more about lots of everyday things, the two-minute anxiety check may fit better, alongside practical steps for breaking an overthinking loop. Social doubts, such as replaying conversations afterwards, work a little differently.

Asking a chatbot for reassurance isn’t a failing. It’s an understandable way to look for relief from a painful doubt, using the most convenient tool there is. The relief just doesn’t last, and the way out is learning to let some questions stay open.