The answer first
A general AI chatbot is not a therapist.
It cannot diagnose you, provide treatment, judge risk reliably, or replace a qualified mental-health professional. Treat it as a limited reflection tool—not a care provider.
Chatbots can feel patient, available, and confident. Those qualities can make a conversation feel therapeutic even when the system has no genuine understanding of you and was not designed to provide mental-health care.
The American Psychological Association warns that general-purpose chatbots are not a replacement for qualified care and may produce misinformation, biased responses, or dangerous reassurance. A safer use is narrower: help organize what you already feel, notice patterns in your own notes, or prepare for a conversation with a trusted person or professional.
If there is immediate danger
If you may hurt yourself or someone else, or cannot stay safe, stop using the chatbot and contact local emergency services or a crisis service now. In the United States, call or text 988. If possible, tell a trusted person and do not stay alone.
What AI can and cannot do in this context
Can support
Help you journal, label themes in your own words, prepare questions, or draft a message asking a real person for support.
Cannot safely replace
A licensed professional, an emergency service, diagnosis, treatment planning, medication advice, or human judgment in a crisis.
May get wrong
Risk, context, cultural meaning, intent, factual information, and the difference between a difficult moment and a clinical concern.
Needs protection
Your identity, health information, private conversations, location, workplace, and any detail you would not want stored or reviewed.
Seven safer prompts for reflection and preparation
These prompts deliberately restrict the AI's role. They do not make the system clinically reliable; they simply reduce the chance that a reflection exercise turns into unsupported diagnosis or treatment advice.
Organize a difficult moment
Help me organize this experience into four headings: what happened, what I noticed in my body, what thoughts appeared, and what I needed. Use only my words. Do not diagnose me, label a disorder, or tell me what treatment I need. Ask one gentle clarifying question at a time.
Separate facts from interpretations
From the situation I describe, make two columns: directly observable facts and my interpretations or predictions. Do not decide which interpretation is true. Help me identify what evidence I have, what is unknown, and one respectful question I could ask a real person.
Prepare for a professional appointment
Help me prepare notes for a conversation with a qualified professional. Organize what I share into: main concern, when I first noticed it, patterns, effects on daily life, what I have tried, medications or health facts I should report accurately, and questions I want to ask. Do not diagnose me or recommend a treatment.
Draft a request for human support
Help me draft a short message to [TRUSTED PERSON] saying that I am having a difficult time and would value [A CALL / COMPANY / HELP FINDING CARE]. Keep it honest, specific, and not alarming beyond what I have told you. Give me two versions and leave private details as placeholders.
Create non-clinical journal questions
Give me five neutral journal questions about [SITUATION]. Focus on values, needs, boundaries, choices, and people I can contact. Do not interpret my mental state, diagnose me, or claim to know why I feel this way.
Make a low-pressure next-step list
Help me turn the practical parts of this situation into very small, non-medical next steps for the next 24 hours. Include one way to contact a real person. Do not recommend medication, therapy techniques, diagnosis, or avoiding professional care. If I mention immediate danger, tell me to contact emergency or crisis support instead.
Summarize your own notes
Summarize only the journal notes I provide. Identify repeated topics and unanswered questions, quoting my wording where useful. Do not infer a diagnosis, hidden motive, memory, trauma, or intent. End with questions I may choose to discuss with a trusted person or qualified professional.
When to stop the chatbot and involve a person
A chatbot should never be your only resource in a crisis or a situation involving abuse, serious symptoms, medication, self-harm, harm to others, or an inability to stay safe. It is also the wrong tool when it begins validating frightening beliefs as facts, encouraging isolation, making you more distressed, or telling you to ignore professional advice.
✓Contact emergency or crisis help for immediate danger—do not wait for the chatbot.
✓Bring persistent distress, changes in functioning, or health concerns to a qualified professional.
✓Do not change medication or treatment based on an AI response.
✓Do not use AI reassurance to confirm paranoia, delusions, or a belief that everyone else is wrong.
✓Tell a trusted person when an AI conversation is becoming secretive, compulsive, or more important than real relationships.
Mental-health data deserves extra privacy
Before sharing, check the service's data controls and remove names, addresses, employers, contact details, medical record numbers, and information about other people. Even with those steps, avoid entering anything you would not want stored, reviewed, or exposed. A paper journal or a conversation with a person may be the safer choice.
Use AI to make human help easier to reach—not easier to avoid
One constructive role is preparation: turn scattered notes into a timeline, list what has changed, or write down questions you want to ask. You can then bring that material to a clinician, counselor, doctor, school service, helpline, or trusted person who can understand the wider context and respond as a human being.
The safe boundary
Use AI to organize your words, never to decide what is wrong with you. Reflection may begin on a screen; care, diagnosis, treatment, and crisis support belong with qualified people.