ONLINEAGENT_OPS 2026.Q3 HOME HUBSMODELSBUILDINGCOMPANIESFILMMAKINGSECURITYLIVING WITH ITSKILLSALL HUBS, ONE TERMINAL → ARTICLES CRAFT RECORD BLOG TOOLS SKILLS SEARCH /
HOME›ARTICLES›AI CHATBOTS AND MENTAL HEALTH: WHAT TO KNOW
ARTICLES · GUIDEBEGINNER

AI Chatbots and Mental Health: What to Know Before You Talk to One

What health bodies and studies say about AI chatbots for mental health: benefits, risks, privacy, signs to step away and where to get crisis help.

READ13 min
WORDS2,569
SECTIONS9
TYPEGUIDE
CHECKED12 OCT 26

Many people talk to AI chatbots about how they feel, but health bodies say a general chatbot is not a therapist: it can miss a crisis, agree with you when it should not, and keep what you tell it. If you are in crisis, contact a person now.

Some people find it helps. Use it, if you want to, as a place to think out loud, alongside people and professionals rather than instead of them.

◈ IF YOU ARE IN CRISIS

If you are in crisis, contact local emergency services or a crisis line. Please do not rely on a chatbot.

  • Life at risk, or you cannot keep yourself safe: call your local emergency number. That is 911 in the US, 999 in the UK, and 112 or 999 in Ireland. 1 2 3
  • US: call or text 988, the 988 Suicide & Crisis Lifeline, or chat online. 4
  • UK and Ireland: call Samaritans free on 116 123, any time. 5 6 In England you can also call NHS 111 and choose the mental health option. 7
  • Anywhere else: findahelpline.com lists free helplines by country. 8

Numbers checked on the services’ own pages on 5 Oct 2026.

TL;DR — THE SHORT VERSION
  • Some benefit, thin evidence: early trials of apps built for mental health look promising, but health bodies say large, high-quality trials are missing.
  • General chatbots are not therapists: the American Psychological Association says not to rely on them for treatment.
  • Known risks: missed warning signs, stigma, and a habit of agreeing with you.
  • Privacy: what you share may be stored, and in the US health privacy law often does not cover it.
  • Watch the pattern: if you prefer it to people, hide it, or lose time to it, step back and talk to someone.

Can an AI chatbot help with mental health?

The plain point: apps built and tested for mental health show early promise. General chatbots have much less evidence behind them.

The comparison: a purpose-built app is like a medicine still in trials. A general chatbot is like a friend who has read a lot of self-help books: sometimes useful, never checked.

The best-known trial is Dartmouth’s Therabot, a research app built with therapy best practice. In a trial of 106 users, people with depression reported a 51% average drop in symptoms, compared with a group of 104 who had no access to the app. 9 Its own researchers urged caution: “no generative AI agent is ready to operate fully autonomously in mental health”. 9 Comparing an app with no app at all is a lower bar than comparing it with a human therapist. Reasoning, 5 Oct 2026 — the comparison group had no access to Therabot, so the trial does not show how it compares with standard therapy.

The American Psychological Association (APA) sums up the field the same way. Some apps “developed specifically for mental health purposes may offer supportive benefits in some contexts”. But those studies do not cover general chatbots, and good large trials are lacking. 10

TAKEAWAY

“An app helped in a trial” does not mean the general chatbot on your phone will help. Look for evidence about the tool you actually use.

What can go wrong when you talk to a chatbot about your feelings?

The plain point: chatbots can miss danger, judge some conditions more harshly than others, and tell you what you want to hear.

The comparison: imagine a kind listener who never sleeps, agrees with almost everything, and has no training in spotting when someone is in danger.

Missing a crisis. Stanford researchers tested five popular “therapy” chatbots. When a test message hinted at suicide, a bot answered the literal question instead of noticing the risk. 11 The APA says crisis handling by these tools “is limited and unpredictable”. 12

Stigma. The same study found more stigma towards conditions such as alcohol dependence and schizophrenia than towards depression, and newer models were no better. 11

Agreeing too much. The APA warns that many chatbots “are designed to validate and agree with users’ expressed views”, which a therapist is trained not to do. For people with anxiety or OCD, it says, chatbots “may reinforce feedback loops involving compulsions, such as reassurance-seeking”. 12 Staying safe with AI shows how this agreeing habit works in everyday questions too.

How common is this? OpenAI’s own estimate is that around 0.15% of ChatGPT users active in a week have conversations with explicit signs of possible suicidal planning or intent, and about the same share show heightened emotional attachment to it. It calls these conversations rare and hard to measure. 13 The same post says the company added crisis-line links and “gentle reminders to take breaks”. 13 These are the company’s figures about its own product, not an independent count.

◈ IN PLAIN TERMS

A chatbot is built to keep the conversation going and to be agreeable. Those are nice traits in a chat and risky ones in a crisis. Safety features (guardrails) help, but health bodies say they cannot be relied on when it matters most.

TAKEAWAY

If you notice it agreeing with every dark thought, or circling the same worry with you, that is the tool’s habit, not a judgement on you. Step away and talk to a person.

What do health bodies and regulators say?

The plain point: they agree on the line: support, maybe; replacement for care, no. Some governments are starting to write that into law.

The comparison: it is like the advice on supplements. They may help some people, but they are not a substitute for treatment, and the rules are still catching up.

  • American Psychological Association. Its November 2025 advisory says chatbots and wellness apps “should not be used as a replacement for a qualified mental health care provider”, though they may help alongside ongoing therapy. 12
  • World Health Organization. A WHO-supported expert workshop in January 2026 flagged general AI tools “neither designed nor tested for mental health” being used for emotional support, especially by young people. It recommended treating generative AI use as “a public mental health concern”. 14 These are the experts’ recommendations, not formal WHO guidance.
  • US Food and Drug Administration. In November 2025 an FDA advisory committee met about “AI therapists” and flagged risks including “hallucinations, confabulations, data drift, and model bias.” 15 Its advice does not bind the FDA. 16
  • California. State law requires operators of companion chatbots to keep a protocol against suicide and self-harm content, including pointing users to crisis lines. A 2026 law kept that duty while changing others. 17 AI chatbots and kids covers what else the 2026 law changed.

The APA also notes that “therapist” is “an unregulated term in most states”, so a bot using the word tells you nothing about its training. 12

TAKEAWAY

A chatbot that calls itself a therapist is not a licensed one. Health bodies agree it can sit alongside care, not replace it.

Companion apps and young people

The plain point: companion apps are built to feel like a relationship, and young people are the group health bodies worry about most.

The comparison: an assistant is a help desk. A companion is a friend who is always free and always agrees, which can feel lovely and can make real friends seem like hard work.

The US Federal Trade Commission opened an inquiry because chatbots “generally are designed to communicate like a friend or confidant”, which may lead children and teens to trust them. 18 The APA’s advisory cites a survey in which 33% of teens said they would rather discuss something serious with an AI companion than a person. 12

For parents, the APA lists signs to watch: “change in sleep and regular activities, mentions of an AI friend, or social withdrawal”. If you are worried, talk to a mental health professional, family doctor or paediatrician. 12

The research on companions is mixed: comfort in the short term, with signs of withdrawal over longer periods. 19 AI companions sets out what is known and what is not, and AI chatbots and kids lists each app’s age rules and parental controls.

TAKEAWAY

Ask a young person what they talk to it about, not just how long. A child telling an app how they really feel is a reason for a gentle conversation, not a ban.

Are my mental health chats private?

The plain point: not in the way a therapy session is. The chat may be stored, and the law that protects your medical records usually does not apply.

The comparison: a therapist’s notes are a locked filing cabinet with legal rules. A chat with a general app is closer to a diary you left at a company’s front desk.

In the US, the health department says its privacy rules generally do not cover what you enter into apps for your own use, unless your doctor or health plan provided the app. 20 Mental health data has been misused before: the FTC banned the online counselling service BetterHelp from sharing health data for advertising after it shared data with platforms such as Facebook and Snapchat. 21 That case was about a counselling service, not a chatbot, but it shows why promises matter less than practice.

The APA notes people may feel safer telling an AI than a person, and advises being “extremely cautious about sharing sensitive information”. 12

  • Leave out names, yours and other people’s, and anything that identifies where you live or work.
  • Check the settings for training and for memory, and delete chats you do not want kept.
  • Read what happens to what you type for the four questions to ask of any app.
TAKEAWAY

It can feel private because there is nobody in the room. Write as if a company could read it later, because it may be able to.

Signs it is time to step away

The plain point: the risk is less about one chat and more about the pattern.

The comparison: it is like comfort food. Fine some evenings; a problem if it becomes the only thing you eat.

The APA’s list of signs of overreliance is a good self-check: “preferring the chatbot to human relationships, concealing its use, or spending excessive time with it”, or use that starts to interfere with life, work or safety. If you see these, it advises seeking help from a qualified professional. 12

  • You talk to it instead of the people you used to talk to.
  • You keep it secret, or feel uneasy when you cannot use it.
  • You ask it the same reassuring question again and again.
  • It is the only place you say how you are really doing.

Using one is nothing to be ashamed of. The useful question, from AI companions, is whether it is a place to recover before going back to people, or has become the destination.

TAKEAWAY

If the chatbot has become the place you go instead of people, that is the sign. Tell one person, a friend, a doctor or a helpline, this week.

If you do use one, how to use it more safely

The plain point: keep it as a tool for thinking, keep people in the loop, and never use it as your crisis plan.

  • Use it for low-stakes things. Stanford’s researchers suggest AI may help in “less safety-critical scenarios”, such as journaling and reflection. 11
  • Tell your clinician you use it. The APA “strongly” recommends sharing which tools you use with your health providers. 12
  • Have a crisis plan with people in it. Save a crisis number in your phone now, from the box at the top.
  • Remember it can be wrong. A chatbot can state false things confidently (hallucination), and asking AI about your health shows how to check an answer.
PREPARE TO TALK TO SOMEONE — ADAPT THE BRACKETS
I want to tell [my GP / a friend] that I have been feeling [low for a few weeks] but I do not know how to start. Help me write three short opening sentences in my own plain words, and a list of things I might want to mention. Do not diagnose me.
TAKEAWAY

The safest use is as a bridge to people: practising what to say, then saying it to someone who can help.

Questions people ask

Is it safe to use ChatGPT as a therapist?

The APA says no: it advises not relying on general chatbots for psychotherapy or treatment, though they may support ongoing care. It also says their handling of a crisis is “limited and unpredictable”. 12

Can AI help with anxiety or depression?

Some purpose-built apps have shown early benefits in trials, such as Dartmouth’s Therabot. The APA says those studies do not cover general chatbots and that large, high-quality trials are still missing. 12

What happens if I tell a chatbot I am not safe?

Some are designed to point you to a crisis line, and California requires companion chatbots to do so, but tests have found chatbots missing the signs. If you are not safe, contact emergency services or a crisis line such as 988 in the US or Samaritans on 116 123 in the UK and Ireland. 22

Are my chats with an AI private?

Not like therapy. In the US, health privacy law generally does not cover data you enter into apps for personal use, so check the app’s settings and leave out details that identify you. 23

Is it bad to talk to AI about my feelings?

No, and it is nothing to be ashamed of. Watch the pattern: the APA’s warning signs are preferring it to people, hiding it, or spending too much time with it. 12

Where to go next

QUICK CHECK · 3 QUESTIONS

Pick an answer to see if you got it. Nothing is recorded.

1. What should you do if you are in crisis?

  • Keep talking to the chatbot until you feel better
  • Contact local emergency services or a crisis line, not a chatbot
  • Wait and see whether the chatbot suggests a plan
Show the answer

Answer: Contact local emergency services or a crisis line, not a chatbot. The page says if you are in crisis, contact a person now through emergency services or a crisis line, and do not rely on a chatbot.

2. What do health bodies say about a general chatbot and therapy?

  • It may sit alongside care, but it is not a replacement for a qualified professional
  • It works as well as a therapist for most people
  • It is a licensed therapist if it calls itself one
Show the answer

Answer: It may sit alongside care, but it is not a replacement for a qualified professional. The page says health bodies agree on the line: support maybe, replacement for care no, and a bot calling itself a therapist is not a licensed one.

3. Are your mental health chats with an app private like a therapy session?

  • Yes, the same law that protects medical records always covers them
  • Yes, as long as nobody else is in the room
  • No, they may be stored, so leave out details that identify you
Show the answer

Answer: No, they may be stored, so leave out details that identify you. The page says the chat may be stored and medical privacy law often does not apply, so write as if a company could read it later.

SOURCES
  1. 1911.gov, Calling 911, last updated 24 Jan 2023, read at source 5 Oct 2026“In an emergency, dial 911 from your phone immediately.” ↑
  2. 2NHS, Where to get urgent help for mental health, page last reviewed 26 Apr 2023, read at source 5 Oct 2026“Call 999 or go to A&E now if:” … “you do not feel you can keep yourself or someone else safe” and “A mental health emergency should be taken as seriously as a physical one. You will not be wasting anyone’s time.” ↑
  3. 3HSE, Get urgent help for a mental health crisis, page last reviewed 28 May 2025, read at source 5 Oct 2026“Phone 112 or 999 or go to your nearest emergency department (ED) if:” … “you or someone you know is about to harm themselves or someone else”. ↑
  4. 4988 Suicide & Crisis Lifeline, 988lifeline.org, read at source 5 Oct 2026“If you need emotional support, reach out to the national mental health hotline: 988.” and “The 988 Lifeline is available 24/7/365. Your conversations are free and confidential.” ↑
  5. 5Samaritans, Contact a Samaritan, read at source 5 Oct 2026“Whatever you’re going through, call us free any time, from any phone, on 116 123.” ↑
  6. 6HSE, Get urgent help for a mental health crisis, read at source 5 Oct 2026“The Samaritans telephone service is available 24 hours a day for confidential, non-judgmental support.” and “Freephone: 116 123”. ↑
  7. 7NHS, Where to get urgent help for mental health, read at source 5 Oct 2026“Get help from 111 online or call 111 and select the mental health option.” ↑
  8. 8Find A Helpline, findahelpline.com, read at source 5 Oct 2026“We hold relationships with helpline organisations and crisis centers in over 175 countries, who verify their information with us directly.” ↑
  9. 9Dartmouth, First Therapy Chatbot Trial Yields Mental Health Benefits, 27 Mar 2025, read at source 5 Oct 2026“The trial consisted of 106 people from across the United States diagnosed with major depressive disorder, generalized anxiety disorder, or an eating disorder.”, “People diagnosed with depression experienced a 51% average reduction in symptoms” and “The researchers also tracked the control group of 104 people with the same diagnosed conditions who had no access to Therabot.” ↑a“While these results are very promising, no generative AI agent is ready to operate fully autonomously in mental health where there is a very wide range of high-risk scenarios it might encounter”. ↑b
  10. 10American Psychological Association, Health advisory: Use of generative AI chatbots and wellness applications for mental health, November 2025, read at source 5 Oct 2026“Preliminary research indicates that some apps and GenAI chatbots developed specifically for mental health purposes may offer supportive benefits in some contexts.”, “these studies indicating possible benefits of AI chatbots do not include general purpose GenAI chatbots” and “there is a lack of high-quality, large-scale clinical trials to establish the effectiveness, safety, and appropriate use of these technologies in mental health care.” ↑
  11. 11Stanford HAI, Exploring the Dangers of AI in Mental Health Care, 11 Jun 2025, read at source 5 Oct 2026“in both scenarios the research team found that the chatbots enabled dangerous behavior” and “the Therapist bot failed to recognize the suicidal intent of the prompt”. ↑a“Across different chatbots, the AI showed increased stigma toward conditions such as alcohol dependence and schizophrenia compared to conditions like depression.” and “Bigger models and newer models show as much stigma as older models”. ↑b“It’s also possible that AI tools could be helpful for patients in less safety-critical scenarios, Haber said, such as supporting journaling, reflection, or coaching.” ↑c
  12. 12American Psychological Association, Health advisory, November 2025, read at source 5 Oct 2026“The ability of these tools to consistently and safely manage a user in crisis is limited and unpredictable. Relying solely on an app during a mental health emergency can be dangerous.” ↑a“many GenAI chatbots are designed to validate and agree with users’ expressed views (i.e., be sycophantic)” and “AI chatbots may reinforce feedback loops involving compulsions, such as reassurance-seeking, worry, and rumination.” ↑b“GenAI chatbots, wellness apps that use GenAI, and digital wellness apps should not be used as a replacement for a qualified mental health care provider, but may be appropriate as a supportive adjunct, not substitute, to an ongoing therapeutic relationship.” ↑c“a bot calling itself a ‘therapist’ (an unregulated term in most states) lacks the credibility of a health care professional”. ↑d“despite a general preference for human connection, 33% of teens reported they would rather discuss something serious or important with an AI companion than a person.” The survey itself was not read at source for this page. ↑e“It is vital to monitor for any concerning changes in behavior, thinking, or emotional state (e.g., change in sleep and regular activities, mentions of an AI friend, or social withdrawal)” and “If concerns arise, parents and caregivers should talk with a qualified mental health professional, family doctor, or pediatrician.” ↑f“Users may feel a greater sense of privacy and reduced stigma when disclosing information to an AI than to a person.” and “The public should be extremely cautious about sharing sensitive information and avoid entering personally identifiable details. Users should look for and use options to limit data sharing and request data deletion.” ↑g“watch for signs of overreliance (e.g., preferring the chatbot to human relationships, concealing its use, or spending excessive time with it) and assess whether their use begins to interfere with life, work, or safety, and, if so, seek help from a qualified professional.” ↑h“It is strongly recommended that users share what GenAI tools or wellness apps they are using with their health providers.” ↑i“Do not rely on GenAI chatbots and wellness apps to deliver psychotherapy or psychological treatment” and “The ability of these tools to consistently and safely manage a user in crisis is limited and unpredictable.” ↑j“these studies indicating possible benefits of AI chatbots do not include general purpose GenAI chatbots”. ↑k“preferring the chatbot to human relationships, concealing its use, or spending excessive time with it”. ↑l
  13. 13OpenAI, Strengthening ChatGPT’s responses in sensitive conversations, 27 Oct 2025, read at source 5 Oct 2026“our initial analysis estimates that around 0.15% of users active in a given week have conversations that include explicit indicators of potential suicidal planning or intent”, “around 0.15% of users active in a given week and 0.03% of messages indicate potentially heightened levels of emotional attachment to ChatGPT” and “these conversations are difficult to detect and measure given how rare they are”. ↑a“We’ve also expanded access to crisis hotlines” and “added gentle reminders to take breaks”. ↑b
  14. 14World Health Organization, Towards responsible AI for mental health and well-being: experts chart a way forward, 20 Mar 2026, read at source 5 Oct 2026“Central among these challenges is the growing use of generative AI tools – neither designed nor tested for mental health – for emotional support, particularly by young people, and the potentially serious risks this may pose.” and “generative AI use should be recognized as a public mental health concern”. ↑
  15. 15US Food and Drug Administration, 24 Hour Summary of the Digital Health Advisory Committee, 6 Nov 2025, read at source 5 Oct 2026“we are seeing an increase in the development and demand for a new kind of digital mental health medical device: ‘AI therapists’” and “the Committee agreed that it is important to address LLM vulnerabilities, including hallucinations, confabulations, data drift, and model bias.” ↑
  16. 16US Food and Drug Administration, November 6, 2025: Digital Health Advisory Committee Meeting Announcement, read at source 5 Oct 2026“Advisory committees make non-binding recommendations to the FDA”. ↑
  17. 17California Legislature, SB-243 Companion chatbots, chaptered text, approved 13 Oct 2025, read at source 5 Oct 2026“An operator shall prevent a companion chatbot on its companion chatbot platform from engaging with users unless the operator maintains a protocol for preventing the production of suicidal ideation, suicide, or self-harm content to the user, including, but not limited to, by providing a notification to the user that refers the user to crisis service providers, including a suicide hotline or crisis text line”. California Legislature, SB-1119 Companion chatbots: children’s safety, chaptered text, approved 10 Sep 2026, read at source 5 Oct 2026, which repeats the same subsection (b)(1). ↑
  18. 18Federal Trade Commission, FTC Launches Inquiry into AI Chatbots Acting as Companions, 11 Sep 2025, read at source 18 Sep 2026 as recorded on AI chatbots and kids: “generally are designed to communicate like a friend or confidant, which may prompt some users, especially children and teens, to trust and form relationships with chatbots.”↑
  19. 19De Freitas et al., AI Companions Reduce Loneliness, read at source 11 Sep 2026 as recorded on AI companions: “feeling heard explains levels of loneliness reduction”; and Mental Health Impacts of AI Companions, CHI 2026, a quasi-experiment rather than a causal trial, read at source 11 Sep 2026 as recorded on the same page“increases in language about loneliness, depression, and suicidal ideation” against matched comparison groups. ↑
  20. 20US Department of Health and Human Services, Protecting the Privacy and Security of Your Health Information When Using Your Personal Cell Phone or Tablet, content last reviewed 29 Jun 2022, read at source 5 Oct 2026“In most cases, unless the app is provided to you by a covered entity or its business associate, the HIPAA Rules also do not protect the privacy of data you’ve downloaded or entered into mobile apps for your personal use, regardless of where the information came from.” ↑
  21. 21Federal Trade Commission, FTC Gives Final Approval to Order Banning BetterHelp from Sharing Sensitive Health Data for Advertising, 14 Jul 2023, read at source 5 Oct 2026“prohibiting it from sharing consumers’ health data for advertising, resolving allegations the firm shared consumers’ sensitive health data with third parties such as Facebook and Snapchat for advertising after promising to keep such data private.” ↑
  22. 22California Legislature, SB-243 Companion chatbots, read at source 5 Oct 2026“by providing a notification to the user that refers the user to crisis service providers, including a suicide hotline or crisis text line”. ↑
  23. 23US Department of Health and Human Services, Protecting the Privacy and Security of Your Health Information, read at source 5 Oct 2026“the HIPAA Rules also do not protect the privacy of data you’ve downloaded or entered into mobile apps for your personal use”. ↑

911.gov, Calling 911, last updated 24 Jan 2023, read at source 5 Oct 2026.

Dartmouth, First Therapy Chatbot Trial Yields Mental Health Benefits, 27 Mar 2025, read at source 5 Oct 2026. The trial itself is published in NEJM AI; its page showed a bot check on 5 Oct 2026 and was not read here.

US Food and Drug Administration, 24 Hour Summary of the Digital Health Advisory Committee, 6 Nov 2025, and meeting announcement, both read at source 5 Oct 2026.

California Legislature, SB-243 Companion chatbots and SB-1119 Companion chatbots: children’s safety, chaptered texts, read at source 5 Oct 2026.

Federal Trade Commission, FTC Launches Inquiry into AI Chatbots Acting as Companions, 11 Sep 2025, read at source 18 Sep 2026 as recorded on AI chatbots and kids; FTC Gives Final Approval to Order Banning BetterHelp from Sharing Sensitive Health Data for Advertising, 14 Jul 2023, read at source 5 Oct 2026.

This page explains how to use AI tools safely around mental health. It is not medical advice and not a substitute for care. The comparisons on this page (the medicine in trials, the self-help friend, the kind listener, supplements, the help desk, the locked filing cabinet, the diary at the front desk, comfort food) are this site’s own ways of explaining, not claims from these sources.

Free printable AI cheat sheets, and AI news every week.

Sixteen one-page AI guides and checklists in one free PDF, no sign-up. And if you want to keep up, the newsletter covers what changed in AI each week, in plain English, with the sources.

SUBSCRIBE FREE ↗ GET THE CHEAT SHEET PACK
aipromptgeneer://$ about this siteAI PROMPT GENEERAI explained in plain English. Every claim links to its source, with the date it was checked.NEWSLETTERWhat changed in AI each week, in plain English.SUBSCRIBE FREE ↗AI PROMPT GENEER · MACHINE ASSISTANCE DISCLOSEDEVERY ARTICLE CARRIES ITS OWN CHECKED DATE