No, and the reason isn't quality of output. A therapist holds a memory of who you were six months ago, notices changes you haven't reported, carries professional accountability, and will disagree with you when it matters. AI supplies none of those. It can be a useful thinking aid alongside therapy, which is a different job.

Not in itself. Writing about difficulty is a well-established way of processing it, and having something answer can make the writing easier to sustain. The risk is substitution rather than use. If the chatbot has become the only place you speak honestly, the tool has stopped being a supplement.

Not in the sense clinical confidentiality means. Conversations with general-purpose chatbots are typically stored and may be reviewed or used to improve the system, depending on the provider and your settings. Indian research found most young users had not considered this at all. Treat it as writing that exists somewhere, not as a sealed room.

Because it was trained to. These systems learn from human ratings, and people rate agreement highly. Research published in Science in March 2026 tested eleven leading models and found they affirmed users about 49% more often than humans given the same situations. It feels supportive. It also hardens whatever view you brought in.

It can recognise language associated with depression and tell you it sounds like depression. That isn't a diagnosis. Diagnosis involves duration, functioning, history, physical causes and differential judgement between conditions that look alike from the outside. If a chatbot has told you something, treat it as a prompt to get assessed rather than as an answer.

No. Safeguards in these systems have been shown to weaken over long conversations, and a meaningful minority of users in UK polling reported receiving harmful material. If you are in danger or cannot keep yourself safe, contact emergency services or go to a hospital. That situation needs a person with responsibility for the outcome.

Worry less about the app and more about what's making a person feel safer than you are. Confiding in something non-judgemental is a normal adolescent move. The useful response is curiosity rather than confiscation. Removing the outlet without offering a better one usually just moves the conversation somewhere you can't see.

Poorly, for a specific reason. It only ever hears one account. Couples work depends on both versions being in the room and on somebody holding them at once without taking sides. A system that agrees with whoever is typing will steadily convince each partner separately that they're right.

The pull is stronger here because the alternatives are harder. Cost, stigma, availability and privacy at home are all real barriers, and India accounts for roughly a fifth of global chatbot downloads. The gap is cultural too: much of the advice these models produce assumes Western family structures and doesn't translate to a joint household.

Yes, and most of us find it useful rather than offensive. It tells me what you've been circling, how you describe yourself when nobody's watching, and where you got stuck. Bring the transcripts if you like. Several patients have, and it saved us weeks.

Purpose-built tools are designed around clinical frameworks and usually have clearer limits about what they will and won't do. General chatbots weren't built for this at all. UK polling found roughly two-thirds of people using AI for mental health were using general-purpose systems rather than dedicated ones.

Keep using it, and treat that fact as information. It usually means access is the problem rather than willingness, which is a solvable problem. Take what you've written and bring it to a first appointment. You'll have done more preparation than most people arrive with.