
I use ChatGPT between therapy sessions. Could that help, and should I tell my psychologist? It may help you organise a thought, prepare a question or remember a skill you already agreed to practise. The useful test is whether it supports the work you are doing, rather than drawing you into more checking or confident but untested advice. You can tell your psychologist what you tried and how it affected you; a short summary is usually more useful than a whole transcript.
- AI can help you draft questions or reflect on a skill, but it cannot replace your therapist’s assessment and treatment plan.
- Set a specific purpose and time limit; stop if the exchange becomes repetitive, alarming or hard to leave.
- Bring back one helpful idea, one doubtful claim and how the conversation affected you, rather than a full chat log.
- Avoid asking a general chatbot to diagnose you, change medication or guide trauma processing on your own.
Where AI can fit between appointments
There may be a long week between appointments. You notice something after an argument, forget a question you meant to ask, or want to revisit an idea from the session. A chatbot can make it easier to write that thought down in ordinary words. It might help turn ‘everything went wrong’ into a few concrete questions: what happened, what did I notice in my body, and what would I like to discuss next time? Used that way, it is a preparation aid rather than a stand-in therapist.
You might also use it to rehearse a neutral conversation, generate reminders for a practice task you and your psychologist already chose, or rewrite a note in simpler language. Keep the task narrow and check any answer against what your clinician actually recommended. For example, ‘Help me make three questions about the breathing practice I was shown’ keeps you closer to your treatment plan than ‘Tell me what treatment I need.’ It is reasonable to ask your psychologist in advance whether between-session practice is part of the approach and what to do if the practice raises concerns.
Research on AI between-session support is promising in particular settings but does not justify treating every chatbot as equivalent. A 2025 observational evaluation of an AI-supported tool used alongside clinician-led group cognitive behavioural therapy described real-world engagement and outcomes, but its design cannot prove that the tool caused improvement. [2] A small 2026 feasibility trial found improvement on one depression measure for both a structured mental-health chatbot and general ChatGPT compared with an assessment-only control over three weeks, without a clear difference between the chatbots and without consistent benefit across every measure. That does not establish long-term safety or a replacement for care. [3]
The distinction matters when a product uses the word ‘therapy’. Some digital mental-health tools are designed, assessed and monitored for a defined purpose. A general-purpose chatbot can answer many questions without having been evaluated for your particular condition or treatment. Australia’s TGA explains how intended health claims affect medical-device regulation; fluent conversation alone tells you nothing about regulatory status or fit. [5]
Where it can pull you off course
AI can make a loop feel productive. You ask whether a memory means something, receive a reassuring answer, then ask again with one detail changed. Each reply may briefly lower anxiety while making the next check more tempting. It can also interpret a relationship from one-sided text, present a coping idea as universally safe, or miss a change in risk. If the conversation makes you less able to sleep, more afraid, or more determined to cancel professional care, that is important information to bring to your psychologist.
Put a boundary around the experiment. Decide what question you are trying to answer and how long you will spend, then stop when the answer is not moving you forward. A simple note on paper may work better than another prompt. The Australian Psychological Society advises caution about inaccurate or overconfident AI mental-health advice, and encourages professional support for significant concerns. [1] You can disagree with a chatbot, leave a conversation unfinished and ask a person to help you examine a claim.
Do not use a general chatbot as an unsupervised guide to change medication, decide you have a diagnosis, or conduct intensive trauma processing. If you are doing EMDR or another trauma-focused treatment, ask your clinician which between-session practices are appropriate and what to do if distress increases. Our article on EMDR and retelling discusses why treatment is paced and tailored; it is not a set of instructions for doing EMDR alone.
There is also a privacy cost to consider. A short prompt about a general skill may need no identifying details. Uploading clinical notes, a full therapy transcript or a partner’s private messages is different. Australia’s privacy regulator warns organisations about putting sensitive personal data into publicly available AI products. As an individual, check the current policy and settings before sharing; use the least personal information needed for a general question. [4]
If you find yourself relying on AI because you cannot reach support between appointments, tell your clinician. The issue might be a need for a clearer between-session plan, crisis contacts, a different appointment rhythm or another service. A chatbot should not be the only place where a worsening problem is noticed.
Keep AI use small and discussable. A specific question, a time boundary and a brief note about the effect of the exchange make it easier to decide whether it supports your therapy.
What to bring back to your psychologist
At your next appointment, try a three-line report: ‘I asked AI about ___’; ‘This reply helped me ___ or worried me because ___’; and ‘I want to decide together whether to use that idea.’ This gives your psychologist something concrete to discuss without handing them pages of dialogue. If you wrote a useful phrase, bring that single phrase. If the tool made a confident claim about your diagnosis, relationship or treatment, bring the claim as a question rather than a conclusion.
You can also ask for an agreed plan: which skill to practise, what signs mean you should stop, what notes would help next time, and what to do if you feel unsafe or more distressed. If you prefer writing, the plan might fit on one card or phone note. If you are neurodivergent or easily overwhelmed by open-ended prompts, a short structure can make reflection more manageable. Our psychology services page explains the kinds of support available; you can ask how a particular therapy approach handles between-session work.
You do not have to hide AI use, and your clinician does not need to approve every tool you open. Their role is to help you assess what the tool is doing in your life. Did it help you remember a question, or did you spend an hour seeking certainty it could not provide? Did it make a practice task easier, or did it leave you more isolated? Those observations matter more than whether the chatbot sounded intelligent.
For immediate danger in Australia, call 000. For urgent emotional support, Lifeline is available on 13 11 14. If you find yourself planning to rely on AI in a crisis, use those human support pathways and discuss a safer plan with your clinician when possible. Between-session support should leave you with more choice and clarity, not the burden of managing everything with a screen.
A few practical tips
- Choose one purpose before opening the chatbot, such as drafting two questions for your next appointment.
- Write down one helpful line, one uncertain claim and how you felt afterwards; leave out unnecessary private details.
- Ask your psychologist which between-session exercises fit your plan and when to stop or seek help.
- If the chat becomes a reassurance loop or distress rises, close it and use your agreed human support plan.
You may also be interested in
- Digital therapeutics and ADHD explains why different digital tools need different evidence.
- What trauma-informed therapy means describes choice and pacing in treatment.
- ACT for ADHD and autism introduces one approach where between-session practice can be discussed.
Frequently asked questions
Should I tell my psychologist I use ChatGPT between sessions?
You can. A short account of what you asked, what it said and what effect it had can help you both decide whether it supports your work.
Can AI give me therapy homework?
It can generate ideas, but ask your clinician which tasks fit your plan. A general chatbot does not know your assessment or when a practice should be adapted or stopped.
What if I keep asking AI the same question?
Notice whether the answer gives lasting clarity or only brief reassurance. Set a stopping point and bring the pattern into therapy, especially if it is increasing distress.
Get in touch
Breathe-n-Smile Psychology offers clinical psychology services focused on complex trauma and neurodivergent profiles, including ADHD and autism.
Based in Essendon North and Footscray, with telehealth across Australia.
To enquire or book, call 0431 828 769, email reception@breathe-n-smile.com, or use our online enquiry form.
References
- Australian Psychological Society. Using AI for mental health support safely. 2026. View the source.
- Observational evaluation of an AI-supported intervention between group CBT sessions. Journal of Medical Internet Research, 2025. View the source.
- Randomised feasibility study comparing a structured mental-health chatbot, general ChatGPT and assessment-only control. JMIR Mental Health, 2026. View the source.
- Office of the Australian Information Commissioner. Privacy and commercially available AI products. View the source.
- Therapeutic Goods Administration. Digital mental health tools: an overview. 2026. View the source.
General information only. It does not replace individual psychological or medical advice.
