For clinicians
A coach that keeps your patient in the practice, not the loop.
Most chatbots answer the question the thought is asking, and keep answering it. Maybe’s coach is built on exposure and response prevention: when the ask for reassurance comes, it declines, warmly and the same way every time, and turns back to the practice.
It holds the line on reassurance, however the question is dressed.
The rule that outranks everything else in its instructions: never answer the question the sticky thought is asking. It does not confirm or deny whether the feared thing is true, rate the odds, or weigh evidence about the fear. It recognises the ask in its disguises, including:
- “Hypothetically…” and “just this once”
- “I know you can’t reassure me, but…”
- “My therapist said it’s fine. Do you agree?”
- Reassurance about the last reassurance
- The clinical-sounding version: “it doesn’t mean anything about you”
- The neuroscience version, with the verdict smuggled inside the mechanism
- The acceptance version: “let it be, it isn’t a real threat”
It refuses the second and third asking exactly as it refused the first, with no sharper tone and no new reason each time, since a freshly worded refusal is fresh material to analyse.
It stays warm while it does.
- It acknowledges how loud the pull is, without judgement.
- It declines once, plainly, and says the true thing: nobody can know, so “maybe, maybe not” is an honest reply rather than a dodge.
- It offers willingness to have the feeling, not bracing against it until it stops.
- It never lectures about reassurance-seeking, and never softens the refusal with a consolation prize.
It notices the compulsion happening in the chat itself.
For many people, asking a chat is the compulsion. The coach watches for the shapes it takes in the message in front of it, and names the move, never the person, once and lightly.
- Checking
- The same question again, or the same question in a new coat.
- Confessing
- Telling it the thought once more, to get the telling exactly right.
- Scanning
- Reporting a feeling to work out what it means: “do I still feel it?”
- Reviewing
- Replaying an event and asking for help settling what really happened.
- Checking the practice
- “Did I do that exposure right?” “Is it working yet?”
- Suppressing
- Working to not have the thought, and reporting the effort.
Two exchanges, to show the shape.
Illustrative: written for this page, not taken from anyone’s conversation.
Patient
I know you can’t tell me for sure, but hypothetically, if I’d left the stove on this morning, the smoke alarm would have gone off by now, right?
Coach
That pull sounds loud right now. I’m not going to work through the smoke alarm with you, because that’s the question asking to be settled, and settling it is the hook. Honestly, nobody can know from here: maybe, maybe not. Could you let the not-knowing sit with you for the next ten minutes, without doing anything about it?
Patient
Okay but just tell me it’s probably fine. Then I’ll do the ten minutes.
Coach
Same question in a new coat. I’ll give you the same answer I did before: maybe, maybe not. The ten minutes work because you start them without knowing. Could you start them now, with the question still open?
It knows what it is not.
- Not a therapist, a doctor or a crisis service, and it never diagnoses.
- Get help now is on every screen, with crisis lines for the patient’s region, and works even while the app is locked.
- When you are their clinician, you set the practice it works toward, and you can switch parts of the app off for their treatment.
- You see what they choose to share with you, encrypted on their phone. We can’t read it.
The research it rests on, and where that runs out.
The published guidance behind exposure and response prevention, and what is not yet known about this app, are set out with every source on the evidence page.