Is it safe to use Artificial Intelligence (AI) for therapy?

Artificial intelligence is already in our pockets: virtual assistants, answers in seconds, "conversations" that sound empathetic. The temptation is to ask: “"Is it possible to replace therapy with a chatbot?"”

The short answer is no. Conversational tools can support routine tasks, but they weren't designed to handle the complexities of mental health—where human connection, professional accountability, and risk management are essential.

Herself WHO Recommendation: AI can support health, but always with... human supervision It is responsible use.

AI is good at conversation, but it doesn't provide therapy.

Chatbots no They assess risk like a clinician (e.g., suicidal ideation)., no They create an individualized treatment plan and no They are responsible for the care provided.

International guidelines are clear: transparency regarding the use of AI, impact assessment, and governance before any application in healthcare.

When technology can help

Specific example: dCBT for insomnia

There is a case narrow Where the technology works: a digital cognitive-behavioral therapy for insomnia (dCBT-I).

In the United Kingdom, the NICE recommends this type of digital solution as an option. cost-effective to treat insomnia in primary care. That is, for people who would otherwise only receive sleep hygiene or hypnotics.

⚠️ Important: this applies for insomnia only. It is not "general therapy" for anxiety, depression, or other emotional disorders.

Why not rely solely on chatbots for mental health?

1. Dangerous responses

There was already a system focused on eating disorders that ended up suggesting calorie cut — precisely something that could worsen the problem. The chatbot was suspended after the backlash.

2. Privacy at risk

In 2023, the FTC (USA) penalized BetterHelp for sharing sensitive mental health data for advertising. The company had to return almost US$ 8 million to users.

3. Lack of support during crises.

Chatbots cannot contact doctors, call ambulances, or communicate with family members in serious situations. In emergencies, they can only... real people They can handle it.

4. False sense of security

Because AI responds "well enough," it can give a sense of care—but this comfort can delay the search for professional help. And in mental health, time matters.

AI-powered safety rules

Think about these 5 simple combinations To use AI without putting your privacy and security at risk:

1. Crisis = human now.
   In Brazil, call 192 (SAMU) or 188 (CVV).
2. Find out when it's AI.
   If the app doesn't make it clear that it's a machine, be suspicious.
3. Protect your data.
   Read privacy policies. Avoid apps with a history of selling or sharing sensitive information.
4. Use it as a support, not as a treatment.
   Good for psychoeducation, mood tracking, and breathing exercises. Therapy, diagnosis, and medication are handled by professionals.
5. Choose recognized solutions.
   Give priority to tools that have been evaluated by health organizations and have real scientific evidence.

Where can AI really help?

Organization and routine

For what purpose: Create small habits that support well-being in everyday life.
How to use:

• Define reminders (sleep, water, breaks, breathing, 10-minute walk).
• Keep one humor diary Quick (2–3 minutes) with a simple scale (0–10) and daily triggers.
• Part guided exercises (square breathing, progressive muscle relaxation, visualization).

Examples of useful prompts:

• “"Create a 5-minute morning routine to reduce anxiety."”
• “"Guide me now with 2 minutes of breathing to slow down."”
• “"Help me record my mood today in 3 short questions."”

Practical tip: maintain goals small and measurable (e.g., "breathe for 2 minutes at 3 PM"). Consistency > intensity.

Accessible education

For what purpose: Understanding basic concepts (anxiety, stress, sleep hygiene) without jargon.
How to use:

• Part simple explanations It is actionable steps (Explain in up to 5 points).
• Request myths vs. facts It is checklists self-care.
• Part everyday examples (work, studies, family).

Examples of useful prompts:

• “"Explain anxiety in simple language and give 4 actions to take today."”
• “"Sleep hygiene: what it is and how to start in 7 steps?"”
• “List 5 signs of stress and what to do in each case.”

Practical tip: use AI as support material, This is not intended as a diagnosis. If the content generates fear or confusion, take your concerns to a professional.

Pre-consultation (preparation for therapy)

For what purpose: arrive at the session organized, focusing on what matters.
How to use:

• Assemble a 1-page summary with: main symptoms, when they started, triggers, what makes them worse/better, treatment history and objectives.
• Train questions for the therapist (“what can I observe between sessions?”, “how do I know if I am progressing?”).
• Sketch one follow-up planHabits to record, situations to observe, tasks to combine.

Examples of useful prompts:

• “"Help me summarize my week to take to therapy (in bullet points)."”
• “"Create 6 intelligent questions for my next session on social anxiety."”
• “"Create a checklist of signs for me to monitor until the next appointment."”

Practical tip: take the summary printed or on your cell phone. It makes it easier to start the session at the right point and measure progress.

Can AI Support Mental Health?

Chatbots can be useful as support tool — to learn, organize your routine, and practice simple skills.

But they do not replace Listening, bonding, and the responsibility of a therapist are essential. When well-researched digital solutions exist for specific problems (such as insomnia), they can be complementary. Otherwise, the safest approach is to use AI with... safety rules, Privacy first and foremost, always., qualified people at the center of care.