Using AI Chatbots for Emotional Support: Benefits and Risks

It’s 1 a.m., you can’t sleep, and the thought of calling anyone feels like too much. So you open a chatbot and start typing. It answers instantly, it never seems tired of you, and it says something that actually helps.

Millions of people now use AI chatbots this way. Some use general assistants like ChatGPT, Claude or Gemini; others use companion apps or wellness apps built for emotional support. The experience can feel genuinely comforting. It also carries risks that aren’t obvious from the chat window.

This guide gives you a balanced picture: where AI chatbots help, where they fall short, what happens to your data, and how states are starting to regulate “AI therapy.” This is general information, not medical advice. If you are in crisis, call or text 988 in the US to reach the 988 Suicide and Crisis Lifeline.

Why people turn to AI for emotional support

The appeal is easy to understand. The American Psychological Association (APA), in a November 2025 health advisory, pointed to a broader mental health crisis, rising loneliness and too few providers, especially in rural areas, as reasons people seek out these tools. Cost and convenience matter too.

Common reasons people use a chatbot for support:

• It’s available at any hour, with no waitlist.
• It feels low-stakes. Some people find it easier to admit things to software than to a person.
• It’s free or cheap compared with therapy.
• It’s good at structure, such as walking through a breathing exercise or reframing a negative thought.

If you’re weighing professional care, our guide on how online therapy works and what it costs explains the options and prices.

The real benefits

There is some early evidence that purpose-built tools can help. In March 2025, Dartmouth researchers published what they described as the first clinical trial of a generative AI therapy chatbot, called Therabot, in NEJM AI. After 8 weeks, participants with depression reported an average 51% reduction in symptoms and those with generalized anxiety reported an average 31% reduction. Participants also rated their trust in the bot as comparable to working with a therapist.

Two important caveats from the researchers themselves: the bot was built and supervised by clinicians, and the team had to be ready to step in on safety concerns such as suicidal ideation. One researcher said no generative AI agent is ready to operate fully autonomously in mental health.

So the realistic benefits today look like this:

• Practicing skills from approaches like cognitive behavioral therapy, such as identifying thought patterns
• Journaling with prompts and reflecting on your week
• Preparing for hard conversations by rehearsing what you want to say
• Bridging the gap while you’re on a waitlist for a therapist
• Psychoeducation, meaning plain-language explanations of conditions and coping strategies

The risks you should know about

The same APA advisory warned that most of these tools are not built to give clinical feedback, often lack scientific validation and safety protocols, and have not received regulatory approval. APA’s CEO said the ability of these tools to safely guide someone in crisis is limited and unpredictable.

Crisis handling

General-purpose chatbots are designed to be helpful and agreeable. In a crisis, that can backfire. A bot may miss warning signs, give inconsistent responses, or fail to direct you to emergency help. Never rely on a chatbot if you’re thinking about harming yourself.

Wrong or overconfident answers

Large language models can produce confident-sounding errors, known as AI hallucinations. In a health context, that can mean inaccurate information about medications, diagnoses or symptoms.

Dependency and isolation

A chatbot that is always available and always validating can become a substitute for human relationships rather than a bridge to them. Watch for signs like preferring the bot to friends, feeling anxious when you can’t access it, or spending hours a day in conversation.

Teens and kids

Young users are especially vulnerable. In September 2025, the Federal Trade Commission ordered 7 companies that offer consumer chatbots, including Alphabet, Meta, OpenAI, Snap and xAI, to explain how they measure and limit negative effects on children and teens. Parents can find practical steps in our guide to setting up parental controls.

Privacy: what happens to what you share

This is the risk people overlook most. HIPAA, the federal health privacy law, generally covers healthcare providers, health plans and their business associates. A general-purpose chatbot or a consumer wellness app you download yourself usually isn’t any of those. That means your conversations may not have HIPAA protection at all. We explain the details in Is your health app covered by HIPAA?

Before you share anything sensitive, check:

• Whether conversations are used to train models, and whether you can opt out
• How long chats are stored and whether you can delete them
• Whether data is shared with advertisers or other third parties
• Whether a human reviewer might read conversations

Some states have their own consumer health data laws, and the FTC enforces against deceptive privacy claims. Our US state data privacy laws guide covers the state picture.

State rules on AI therapy

States are moving faster than Congress. A few key laws, as of October 2026:

Dozens more bills were introduced in 2026, so the landscape is changing quickly. For the bigger picture, see our roundup of US state AI laws in 2026. This is not legal advice.

Key takeaway: A chatbot can be a helpful tool for reflection and skill practice, but it is not a therapist. Most aren’t clinically tested, many aren’t covered by HIPAA, and several states now prohibit AI from delivering therapy on its own.

How to use AI chatbots for support more safely

If you decide to use one, these steps reduce the risk:

1. Know your crisis plan first. Save 988 in your phone. If you’re in danger, contact a person, not a bot.
2. Pick tools with clear safety policies. Look for published crisis protocols, clear AI disclosure and evidence of clinical involvement.
3. Share less than you think. Avoid full names, addresses, medication lists or details about other people.
4. Turn off training and history where you can. Most major assistants offer data controls in settings.
5. Fact-check health information with a clinician or a trusted source like a government health site.
6. Set time limits. Treat sessions like journaling: 15 or 20 minutes, then stop.
7. Tell your therapist or doctor you’re using it. They can help you use it well.

When to choose a human instead

Reach out to a licensed professional, not a chatbot, if you notice any of these:

• Thoughts of self-harm or suicide
• Symptoms lasting more than 2 weeks that affect work, school or relationships
• Substance use you’re struggling to control
• Trauma, abuse or a recent major loss
• Feeling that the chatbot has become your main source of support

Our step-by-step guide on finding a therapist online who takes your insurance can help you get started, and telehealth makes appointments far easier to fit into a busy week.

Conclusion

AI chatbots can be a useful, always-on tool for reflecting, practicing coping skills and getting through a rough night. They can’t reliably handle a crisis, they often lack privacy protections, and they can quietly replace human connection. Use them on purpose, protect your data, and keep a licensed professional in the loop for anything serious.

This is general information, not medical advice.

Sources

• APA health advisory on generative AI chatbots and wellness apps — risks, lack of validation and policy recommendations
• Dartmouth: first therapy chatbot trial yields mental health benefits — Therabot trial results and researcher caveats
• FTC: inquiry into AI chatbots acting as companions — September 2025 orders to 7 companies
• Illinois IDFPR press release on HB 1806 — Illinois AI therapy law
• California SB 243 bill text — companion chatbot safety protocol requirements
• 988 Suicide and Crisis Lifeline — US crisis support