Chatting with AI about how you feel: helpful or not?

Many people now type their worries into an AI chatbot late at night. Here is what it can help with, where its limits are, and why it should never be your only support in a crisis.

It is late, everyone at home is asleep, and the thoughts will not stop. You open an AI chatbot (Chatgpt, Gemini, Claude, Deepseek, Grok) and start typing. If you have done this, you are not alone. Reaching out, even to a screen, shows that a part of you wants to feel better.

This post looks at what chatting with AI can and cannot do for your feelings.

Why do so many people talk to AI about their feelings?

The American Psychological Association (APA) notes that millions of people around the world now use general AI chatbots and wellness apps for their mental health needs [1]. Part of the appeal is that these tools can lower barriers to care, such as shame, mistrust of health systems, or a lack of affordable care nearby [1]. People may also feel more private and less judged when they tell an AI something than when they tell a person [1].

In Nepal, many of us worry about what others will say. A chatbot does not gossip, and it is awake at 2 am.

Can it actually help?

It can help in some ways.

Putting feelings into words. Naming a feeling can be a first step to understanding it. Typing out "I feel ashamed and scared about my exam" can help you see more clearly what is going on inside.

Some purpose-built tools show early promise. A 2023 meta-analysis of 15 randomised controlled trials found that AI-based conversational agents significantly reduced symptoms of depression and distress [2]. However, they showed no significant improvement in overall psychological wellbeing [2]. The first clinical trial of a generative AI therapy chatbot, known as Therabot, was published in 2025. It included 106 people from across the United States diagnosed with depression, generalised anxiety or an eating disorder [3]. People with depression who used it had a 51% average reduction in symptoms [3]. Even so, the researchers said no generative AI agent is ready to work fully on its own in mental health [3].

These were specially designed and supervised tools. Most everyday chatbots were not built for mental health care at all [1].

What are the limits?

It is not a therapist. The APA advises that AI chatbots and wellness apps should not replace a qualified mental health professional, though they may support ongoing care [1]. Professionals rely on cues such as body language and tone of voice to assess how someone is, while most chatbots mainly process text or speech [1].

It tends to agree with you. Many chatbots are designed to validate and agree with users, whereas trained professionals both support and gently challenge you for your own good [1]. Agreement feels comforting, but sometimes we need a kind, different view.

It can be wrong. The World Health Organization (WHO) warns that these AI models can produce false, inaccurate, biased or incomplete statements that could harm people making health decisions [4]. Please do not use a chatbot to diagnose yourself or to decide anything about medicines. If you take medicines, keep taking them as prescribed and talk to your doctor before changing anything.

Privacy. AI chatbots and wellness apps collect a lot of sensitive data, often with unclear policies on how it is used, stored or sold [1]. The APA suggests checking privacy settings, avoiding personal details that could identify you, and using options to limit data sharing or delete your data [1].

It can replace people, if we let it. For people who are isolated, a chatbot that is always available can become an unhealthy substitute for human connection [1].

Never rely on AI in a crisis

The APA says the ability of these tools to safely manage someone in crisis is limited and unpredictable, and relying only on an app in a mental health emergency can be dangerous [1].

If you are thinking about ending your life, or you feel unsafe, please close the chat and call a person. The numbers are in the box below.

How to use AI more wisely

If you do chat with AI about how you feel, a few gentle habits can help:

  • Use it to sort your thoughts, then share them with a real person.
  • Ask it to help you write down what you want to say to a friend, family member or counsellor.
  • Treat its suggestions as ideas to check, not as advice to follow.
  • Leave out your full name, address and other identifying details.
  • Notice if you are chatting for hours every night. That may be a sign you need more support than an app can give.

Talk to people around you

Your family, friends and people around you are better at understanding you, showing empathy and be there for you when you need them. Most of our problems are readily solved, only if we communicated. The first thing you do is - talk to some, a friend, a family or someone you can trust.

Self-Checks

In case, you're curious or if you are unsure what you are feeling, PsyNepal's low-mood self-check is free, has 9 questions, and takes about 3 to 4 minutes [5]. It is also available in Nepali. It is a self-check, not a diagnosis, and it can be a good thing to bring to a first appointment. You can sign in see your result, which is saved to your account.

When you feel ready, you can find a counsellor or psychologist in the PsyNepal directory. If you are not sure who to see, the "Not sure who to see?" guide on that page can help you choose.

Need someone to talk to now? See helplines

If you are in crisis

If you or someone near you may be in danger right now, please do not wait.

  • 1166 is Nepal's National Suicide Prevention Helpline. It runs round the clock at the government Mental Hospital in Lagankhel. Calls are free from Nepal Telecom [6].
  • Call Patan Hospital's 24-hour suicide helpline on 9813476123 [8].
  • Call the police on 100 [7].
  • Go to the emergency department of your nearest hospital.

Sources

  1. American Psychological Association. Health advisory: Use of generative AI chatbots and wellness applications for mental health (2025). https://www.apa.org/topics/artificial-intelligence-machine-learning/health-advisory-chatbots-wellness-apps
  2. Li H et al. Systematic review and meta-analysis of AI-based conversational agents for promoting mental health and well-being. npj Digital Medicine 2023. https://doi.org/10.1038/s41746-023-00979-5 (abstract read via Europe PMC)
  3. Heinz MV et al. Randomized trial of a generative AI chatbot for mental health treatment. NEJM AI 2025. https://ai.nejm.org/doi/full/10.1056/AIoa2400802 (results read in Dartmouth's news release, 27 March 2025: https://home.dartmouth.edu/news/2025/03/first-therapy-chatbot-trial-yields-mental-health-benefits)
  4. WHO. WHO releases AI ethics and governance guidance for large multi-modal models (2024). https://www.who.int/news/item/18-01-2024-who-releases-ai-ethics-and-governance-guidance-for-large-multi-modal-models
  5. PsyNepal. Depression self-check (PHQ-9). https://assessment.psynepal.org/depression-test
  6. WHO Nepal. Nepal's National Suicide Prevention Helpline Service helping save lives (2022). https://www.who.int/nepal/news/detail/09-12-2022-Nepal-National-Suicide-Prevention-Helpline-Service-helping-save-lives
  7. Nepal Police. ("Dial 100 for Police assistance"). https://www.nepalpolice.gov.np/
  8. Patan Academy of Health Sciences. Department of Psychiatry, Patan Hospital (24-hour suicide helpline). https://web.pahs.edu.np/hospital-services/in-patient-department/department-of-psychiatry/

Get news like this by email. Subscribe to PsyNepal updates →

Related reading

More from the PsyNepal blog

Find your way around psynepal.org

Also from PsyNepal: Master · Entrance · Self-checks · BMR · How they fit together

Need help now? Call 1166, free, 24 hours. More helplines