Korea's guidelines on using generative AI for mental health: 21 recommendations and how to check an AI counseling app

CareOps editorial team · October 13, 2026 · 11 min read

Mental wellness information and service guidance published by CareOps.

Translated from our Korean guide. Korean public services, insurance, and helpline numbers refer to South Korea. For urgent help elsewhere, contact local emergency services or a crisis line in your country.

Telling ChatGPT or Gemini about a worry has become common. In a Korean government survey, nearly seven in ten generative AI users also used AI for emotional purposes such as everyday conversation, comfort, or counseling-style talk. In October 2026, the government released its first guidelines to help people do this more safely.

What are the guidelines on using generative AI for mental health?

They are user guidelines released jointly on October 8, 2026, ahead of Korea's Mental Health Day, by the Ministry of Health and Welfare and the Korean Neuropsychiatric Association. The aim is to reduce the risks that can come with using generative AI for mental health purposes: exposure of personal information, inaccurate information and judgments, and overreliance. After consulting experts in fields including media and engineering, the authors produced two versions: one for all users and one for children, adolescents, and caregivers.

The guidelines draw on a survey by the Korea Institute for Health and Social Affairs. From August 25 to September 4, 2026, it asked 1,019 people aged 15 to 79 who had used generative AI in the past month, online.

The most common reason for confiding in AI was "I can talk right away, anytime and anywhere" (85.6%), and more than half also chose "visiting a clinic or the cost is a burden" (59.2%). Because so many people find it helpful, guidance on using it well became necessary.

What are the three key messages?

1. Look into it and decide before you start talking about your feelings. What you tell an AI cannot be taken back, so first check what kind of service it is and how your information will be used.
2. Treat what AI says about your condition as input rather than a conclusion. AI answers based on what you type. Even when an answer sounds convincing, hold it loosely instead of as a diagnosis of your condition.
3. Keep a healthy distance from AI, and talk to people. The empathy, comfort, and advice people give matter for recovering and staying well.

What do the 21 recommendations for all users say?

Each key message comes with seven recommendations. The lists below are condensed paraphrases; the full wording is in the sources at the end of this article.

Look into it and decide before you start

• Check who the service was made for, what kind of help it offers, and whether there is a basis for its benefits
• If a service claims medical effects, check whether it has been reviewed and approved by an accredited body
• Avoid judging a service as safe or effective just because it feels convenient and kind
• Check how the information you enter is stored and used
• Leave out details that could identify you or someone else, such as names, contact information, and addresses
• Ask only about what you want to know instead of uploading photos or files of medical certificates, prescriptions, or test results
• Decide before you start what you will talk about, how far you will go, and when you will stop

Treat what AI says as input rather than a conclusion

• Remember that diagnosing and assessing your condition is the role of mental health professionals
• Hold any diagnosis or assessment an AI offers loosely rather than accepting it as given
• When an AI agrees with you, avoid treating that as proof that your view is right
• Rather than treating an answer as the only correct one, ask for other perspectives and opposing views as well
• Ask directly why it answered that way and what information it relied on
• Check verified sources separately instead of trusting AI-generated text or images as they are
• Set aside answers that are out of step with your situation or the norms of your community

Keep a healthy distance from AI and talk to people

• When a service imitates a person's appearance, voice, or personality, keep in mind that you are talking with an AI
• If your usage keeps growing or you spend less time talking with people close to you, step back on purpose
• Discuss important decisions with someone you trust or a professional instead of asking only AI
• If a response is risky or inappropriate, stop the conversation and tell someone close to you or a professional
• If you are receiving treatment or counseling, tell your clinician or counselor which AI service you use
• When you need treatment or care, get help from established mental health institutions and professionals
• In a crisis, ask mental health institutions and professionals for help right away

How do you apply the guidelines when choosing an AI counseling app?

The first group of recommendations works as a set of criteria for choosing a service. Before you download an app, look for these five things on its introduction page and in its privacy policy.

How does CareOps Maeumi answer these checklist items?

CareOps is a mental wellness app where you speak or write with Maeumi, an AI companion whose conversations are informed by established counseling approaches. This article was written by CareOps, and the table below is our own description of our service against the checklist. It lists only what is stated on our public pages and in our privacy policy.

For how this differs from general-purpose AI, see ChatGPT vs. AI counseling apps; for criteria on choosing an app, see how to choose a mental wellness app.

What do the guidelines recommend for children, adolescents, and caregivers?

In the survey, 69.2% of adolescents aged 15 to 18 used AI for emotional purposes. While 83.1% of caregivers said they knew about it, only 2.5% had set usage rules together. So the guidelines offer seven recommendations each for young people and for caregivers.

For children and adolescents

• Sort out your own thoughts first rather than simply following what AI says, and check whether a suggestion suits your age
• Keep details that could identify you, such as your name, school, home, contact information, and photos, to yourself
• Remember that you can end a conversation whenever you want
• Check for yourself whether you are treating an AI character like a real friend or partner
• If AI is the only place you open up, try sharing the same story with someone close to you too
• If an AI says something risky or strange, tell an adult you trust instead of handling it alone
• Know in advance where you can ask for help when things are hard, such as a school counselor or a youth counseling service

For caregivers

• When a child starts using AI, set basic rules together about why and how it is used
• Check whether the service was designed for children and adolescents from the start, and whether its privacy settings are at the safest level
• When a child says they use AI, listen first to what they use and how, rather than scolding
• Watch for signs that a child treats an AI character like a real friend or partner and leans on it emotionally
• Check whether the mental health information a child gets from AI is safe, reliable, and appropriate for them
• Watch whether usage keeps growing or the child relies on AI far more than on conversations with people
• If you see signs of a crisis, ask professionals and specialized institutions for help right away

If you already talk to ChatGPT about your worries, what can you change today?

1. Leave out names and contact details. Replace your name, other people's names, company or school names, and addresses with words like "my manager" or "a friend."
2. Set the scope and the ending when you start. Saying "Today I'll talk about just this one thing for 20 minutes and then wrap up" keeps the conversation from running on.
3. Ask for other views and reasons. Try "Tell me how this could be seen the other way" or "Tell me why you think so."
4. Discuss important decisions with people too. Decisions such as quitting a job or ending a relationship deserve more than an AI's answer.
5. Keep a summary of what you have discussed. A summary of your concerns so far makes it easier to continue with someone close to you, a professional, or a counseling app. Our Korean page on continuing a conversation you started with another AI has a sentence you can use to create one.

This article shares general mental wellness information. For urgent help, contact local emergency services or a crisis line in your country. In South Korea, the suicide prevention helpline is 109 (24 hours), and the mental health crisis line is 1577-0199.