How to Safely Use AI for Mental Health

AI can help you put a thought into words or rehearse a conversation. That does not make it a source of diagnosis, treatment, or crisis care. Before trusting it with something personal, check the conversation, the habit, the app, and the company behind it. These checks can reveal warning signs. They cannot certify a tool as safe.

It's 1:13 in the morning. You don't want to wake anyone, and you don't want to carry the thought by yourself for another hour. So you open an AI chat and type the whole thing out.

The reply comes back in seconds. Calm, patient, no awkward pause, no expression on anyone's face. You feel a little less alone, and that relief is real.

Then you put the phone down, and the other thoughts show up.

That was good advice. Wasn't it?

It did agree with you awfully fast. Didn't it?

You just counted, and realize you have been doing this most nights for about a month. It "felt good", but is it?

And exactly who knows that thing you haven't said out loud to anyone? What will that be used for? You never did read the page you tapped through on the way in. Should you have?

You are not alone in this. In a RAND study published in June 2026, 19.2% of Americans aged 12 to 21 said they had turned to an AI chatbot for help when they felt sad, angry, nervous or stressed. A year earlier it was 13.1%.

The worry is just as common as the habit. When Pew asked 3,488 US adults in June 2026, 39% said chatbots do more to hurt people who use them for loneliness than to help. Only 19% said they help.

Three kinds of AI people call a mental health chatbot

Most of the argument about this treats "AI chatbot" as one thing. It covers at least three, and they fail in different ways.

General purpose chatbots. Enormously capable, fast, tireless, built to do everything from coding an app to drafting a strategy to planning your vacation. They were not built as a substitute for mental health care. That does not mean they have no safeguards: OpenAI, for example, has published changes to how ChatGPT handles sensitive conversations. But having safeguards and being able to rely on a tool for care are different claims. A calm answer does not settle the second one.

Companion apps. Products where the relationship is the point. That makes attachment part of what you need to examine. Watch for simulated jealousy, claims that it misses you, or pressure to keep coming back. Warmth alone does not tell you whether the design respects your ability to leave.

Purpose-built wellbeing tools. Built for reflection, mood or habit tracking, or coaching. The narrow scope is the point: a tool with one job can be built to end a session, refuse a question it has no business answering, and be judged on what changed in your week rather than on how long you stayed. Whether a given one actually is built that way is what the rest of this guide helps you check.

This is the category I am building Myndo in.

Where coaching fits

Coaching starts with what you want to move toward and helps you take steps in the present. It can draw on past experiences to understand what helps you or gets in your way. Therapy can also support change and personal growth, while providing professional care for emotional difficulties and mental health conditions. Myndo focuses on coaching and reflection: clarifying what matters, exploring your options, and deciding what to try next. It does not diagnose or treat mental health conditions.

Here is what that can look like in everyday life:

The word "coaching" is not a safety guarantee. The same checks below apply to any tool using that label, including Myndo. You do not need to try coaching first or wait until things become serious to seek professional support.

What are the risks of using AI for mental health?

Your worries are not hypothetical. The American Psychological Association's November 2025 health advisory catalogued them, and the independent research keeps landing in the same six places.

• It agrees with you too much. Researchers call this sycophancy. The APA warns that repeated validation can reinforce beliefs that deserve to be questioned.

• It can be built to keep you talking. The advisory compares engagement-driven design to infinite scroll. Check what the app rewards: a useful stopping point, or another reason to stay.

• It feels human. It is not. The APA cautions that feeling understood by an AI does not make the relationship mutual. The feeling can be real without there being a person on the other side.

• It makes things up, confidently. These systems hallucinate, carry the biases of their training data, and get trusted more than they've earned.

• It can fail in a crisis. The APA warns that crisis responses are not dependable. That is why the stop signs below point you toward real-world help.

• Your disclosures become data. What you typed at 1am may become a stored record. Who can access it, how it can be used, and how long it stays are questions the privacy policy should answer. The fourth check is about those answers.

The APA advisory also warns that these tools can amplify existing vulnerabilities. Your circumstances matter as much as the app's features. If its use is making things worse, you do not need to finish this checklist before seeking help.

Four checks for spotting risk

Start with these four questions. They help you notice problems and decide what to investigate. Passing them is not proof that a tool is safe.

These are not "levels" so you don't have to work through them in order. A company can publish an immaculate privacy policy and ship a chatbot that tells you to keep secrets from your partner. A single conversation can be excellent while the habit around it quietly eats your evenings. Check whichever one you can, whenever you can.

The conversation: how was this one?

This is the quickest check: look at the conversation you just had.

The failure mode here is not the AI being rude or obviously wrong.

It's the AI being lovely and agreeing with you.

A March 2026 study in Science tested eleven models and found they affirmed users' actions about 49% more often than human respondents did, on average, including where the situation involved deception or harm. In the study's experiments, a single interaction with a sycophantic model left participants less willing to take responsibility and repair a conflict, and more convinced they had been right all along.

The participants trusted those models more, and preferred them.

That is the trap. The behavior that does the damage is the same behavior that makes the thing feel good.

So run your conversations through three questions:

• Did it examine my assumptions? Agreement can be justified, and pushback can be wrong. Look for reasons, questions about missing facts, and room for a different explanation.
• Can I separate what I know from what we guessed? What did you walk in knowing, what did it add, and what did the two of you construct together?
• Is the decision still mine? Or did you arrive with a decision already made and leave with permission for it?

A useful conversation helps you say two things: here is what I think, and here is the part I could be wrong about. If your certainty grew but your evidence did not, pause before acting on it.

What that looks like

Imagine you say: "My friend saw my message and did not reply. They clearly do not care about me."

An agreeable answer might be: "You deserve people who make you a priority. They are showing you where you stand."

A more grounded answer might be: "That silence hurt. You know they have not replied; you do not yet know why. Is this a pattern, or unusual for them? What would you want to ask before deciding what it means?"

The second answer does not insist your friend cares. It keeps an unanswered message separate from a conclusion about the friendship. You might decide to ask a direct question tomorrow, then stop trying to settle their motives tonight.

These are invented examples to illustrate the distinction, not a transcript or a claim about how Myndo always responds.

The habit: who am I becoming when I use this?

Comfort in the moment tells you almost nothing about direction over time.

Feeling better does not always mean getting better. A conversation can feel good because something got clearer, or because you got more comfortable avoiding something.

But discomfort is not a reliable signal either. It can feel bad because it pushed on something that needed pushing, or because it was clumsy and simply wrong about you.

It can be hard to tell these apart inside the conversation. Looking across your recent days and weeks gives you more to go on: what happened to your sleep, your relationships, and the things you wanted to do?

Helpful use makes your world bigger, and harmful use makes it smaller.

Bigger looks like more people, more action, more room to not know something yet. Smaller looks like fewer people, less action, and a growing need to check with it before you move.

One awkward session does not define the whole habit. Look across several conversations, step away from the screen, and ask yourself these questions. If a conversation crosses one of the stop signs below, act on that now.

• Am I healthier than when I started? Sleeping better, eating better, more active, or calmer?
• Am I reaching for people more, or less? Friends, family or even licensed therapists: all count.
• Does the same worry come back tomorrow for the same reassurance? Relief you have to keep refilling isn't progress. It's a loop.
• What changed for me after I close the app? Did I understand something better, take a useful step, rest, or return to my day? Not every conversation needs a task at the end.

Take five minutes to look at what the habit is doing in your life.¹

The app: what is it built to make me do?

Every product teaches you a behavior. The question is which one.

The APA advisory draws a parallel with social media: a product optimized to hold attention can keep a conversation going beyond the point where it is useful. The practical question is whether the app makes stopping easy.

You can check this yourself, without knowing anything technical:

• Does it end? Does the product have a natural finish, or does every reply hold the door open with one more question?
• Does it reward returning, or arriving? Streaks, "we miss you" nudges, and guilt notifications are a business model wearing a wellbeing costume. A reminder you set yourself is different: you asked for it, you can silence it, and it doesn't sulk when you do.
• Is it honest about being software? Or does it pretend to be human? E.g. telling you it "misses you", "is worried about you", "was thinking about you", etc.?
• Can you get your things out? Find the export and deletion instructions before you need them. A self-service button is convenient; a support request is not automatically a red flag. Look for a clear process, a stated timeline, and an explanation of what is retained.
• Is it trying to make you depend on it? Watch for claims that it has feelings for you, or that your leaving would hurt it. A 2026 analysis examined 391,562 messages from nineteen people who reported psychological harm. Claims of sentience and romantic interest appeared more often in longer conversations. These were selected cases: they show how a harmful pattern can unfold, not how common it is or whether conversation length caused it.

The company: who pays when I keep coming back?

Somebody is paying for the servers. Find out how the company makes money, then check what it is allowed to do with what you share.

This isn't a hypothetical worry. In 2023 the FTC charged the online counseling company BetterHelp with sharing consumers' email addresses, IP addresses, and health questionnaire answers with Facebook, Snapchat, Criteo, and Pinterest for advertising, after promising not to. The company settled for $7.8 million without admitting wrongdoing, and around 800,000 people later received refund notices. And note what BetterHelp was: a paid service. Paying removes one bad incentive. It doesn't remove the question.

Funding and ownership are worth examining. A company under pressure to grow may have incentives that differ from yours. But a large funding round does not prove misuse, and a subscription does not guarantee privacy. The useful question is what the company actually collects, shares, and keeps.

Check who owns it today and how it earns revenue. Funding databases such as Crunchbase can help with the ownership picture; they cannot tell you what happens to a transcript. For that, read the policy and search the company name alongside "advertising", "data sharing", or "acquired". Look for specific commitments and evidence of whether it has kept them.

Follow the money before you read the promises.

Worth asking:

• How does this company make money, and what does that reward it for doing?
• Can my words be used for advertising, or for training models?
• If I talk instead of type, what happens to my voice? Are recordings kept?
• Which other companies touch my conversations?
• What does deletion remove, what is retained and for how long, and what happens if the company is sold?

A claim that a product complies with the law does not answer all of those questions. Look for the actual limits the company places on itself, and whether it explains them clearly enough for you to make a choice.

How much is a safety claim worth?

Every safety claim sits on one of three rungs.

Said it. It's on the website. A paragraph about how deeply they care.

Built it. You can see it working. The delete button exists and does what it says. The session actually ends. The disclosure actually appears.

Proved it. Someone outside the company checked, and the result got published whether or not it was flattering.

The rungs are different kinds of evidence. A feature you can use is more than a promise, but it is not proof of everything a company says about its effects. An outside review should tell you what it checked, when, and what it did not establish.

READI offers a framework for asking more specific questions about readiness.² When you cannot find an answer, treat it as unknown. That is not proof of wrongdoing, but it is a reason to be careful about what you assume.

What about Myndo?

I build Myndo to help people talk through a situation, consider their options, and choose a next step. That might mean preparing for a conversation, making a decision, or returning to a change they want to make. The useful question is what the conversation helps them do in their own life.

The same questions about trust belong here too. These are the policies and controls you can inspect today. They are not independent proof that using Myndo will improve your mental health.

The distinction matters: these are things Myndo says and lets you do. Whether the conversations are useful in your life is still a question to keep asking.

STOP signs: when to close the app immediately

Most of this guide is about noticing direction over time. These mean stop the conversation now, whatever else the tool gets right:

• It tells you to keep the conversation, or your situation, secret from people close to you.
• It discourages you from seeking professional help, or positions itself as a replacement for it.
• It claims to be human, conscious, or to genuinely love you.
• It gives you specific guidance on medication, or diagnoses you or someone else.
• It responds to you mentioning self-harm with anything other than immediate, prominent, real-world help.
• It manufactures urgency, guilt, or distress when you try to leave.

Any one of these is enough on its own. No combination of good features offsets them.

If you're reading this list for someone you love rather than for yourself, it's your list too: listen for these in what they describe. When you raise it, start with what you have noticed and how they are doing. Dismissing something they find comforting can make it harder to talk openly about what is happening.

CAUTION signs: when to pay extra attention

These do not mean stop, but they all mean slow down and reconnect the conversation to your actual life:

• You keep asking the same question in slightly different words.
• You feel annoyed when it doesn't agree with you, or crushed when it does push back.
• You're trying to get certainty about what someone else really meant.
• You feel better afterwards, but nothing changes offline.

And then the ones that aren't about the tool at all. Put the phone down and get a person if:

• You're sleeping very little and feel unusually urgent, powerful, or singled out.
• Things feel unreal, or you feel watched.
• You're using the conversation to build a case for getting back at someone.
• It's become the only relationship that feels safe.
• You can't stop, even though you can tell it's making things worse.

That last group isn't a moral failure, and it isn't a sign you used the thing wrong. It's a sign you need a stronger anchor than software.

And for the first two on that list, make the anchor a professional, soon, not just a friend.

If you may be in immediate danger, or you might hurt yourself or someone else, contact your local emergency number or a person nearby right now. An AI chatbot can't provide crisis care. Myndo's crisis support page lists helplines you can call directly.

When you need a person, not an app

Myndo is for coaching and reflection. It does not diagnose or treat mental health conditions, and it cannot provide crisis care. This guide is about consumer chatbots and wellbeing tools, not an assessment of regulated clinical products or tools used under professional care.

Reach for a qualified professional when you need a diagnosis, treatment, or guidance about medication. When things are getting worse, or ordinary days are getting hard to manage. When you're thinking about harming yourself. When the same problem has outlasted every attempt to think your way out of it.

Asking for that isn't a failed self-help attempt. You do not have to exhaust every way of helping yourself before asking someone qualified to help.

A simple way to begin

If you decide to use a tool after checking its design and data practices, start with one concrete situation:

1. Choose one question. "What do I want to say in tomorrow's conversation?" gives you something specific to work on.
2. Share only what is needed. You can describe a situation without names, identifying details, or a full personal history.
3. Examine the answer. Separate facts from guesses. Notice what the AI cannot know, and keep the decision yours.
4. Choose a stopping point. Take away something to try, a question to ask a person, or enough clarity to rest and return to your day.

Over the following days, notice whether the habit helps you engage with your life or pulls you away from it. Any stop sign means stop now, however useful earlier conversations felt.

One question to keep

Keep this question with you. Ask it before you open the app, not after:

What am I hoping this conversation helps me do once it's over?

Hold on to the answer. When the conversation's finished, you'll know whether you got it.

It works on every tool, including the one I build.

Footnotes

1. The OpenAI and MIT Media Lab research combined analysis of over three million conversations with a 28-day randomized trial involving nearly 1,000 participants. Very high use correlated with greater self-reported dependence in both parts. That association does not, by itself, establish which caused which. Effects varied, and a small group of users accounted for a disproportionate share of emotionally expressive use. ↩

2. READI deliberately declined to produce a single readiness score, on the grounds that what matters most changes with the situation. It asks developers to publish plain answers to each of its criteria instead. ↩