How to use ChatGPT as a life coach
A prompt to start
Paste this into a new chat. It sets up the four things the research below supports: one change, questions before advice, a plan with a fallback, and a check-in that changes the plan rather than judging you.
You're my coach for one change this week. Ask me questions before you give advice, one at a time. Help me pick one thing to do or stop, small enough that I'd still do it on a bad day. Write the plan back to me in my own words: what, when, where, and what I'll do if something gets in the way. Each time I come back, ask how it went before suggesting anything. If I missed, ask what got in the way and change the plan, not your view of me.
Make the plan small and specific
Say what you will do, when and where: “after breakfast, I walk for 20 minutes”. Plans in that if-then form have a real but small effect, d = .15 across 642 tests once publication bias is allowed for (Sheeran, Listrom & Gollwitzer, 2025). Doing it at the same moment in the same place is what lets it become a habit, because habits are cued by context rather than by a goal you keep in mind (Wood & Neal, 2007).
Decide what you’ll do when it gets in the way
Ask yourself what is most likely to stop you, and what you will do the moment it happens. That is coping planning. In a study of 211 people in cardiac rehab it mattered at follow-up rather than straight away (Sniehotta, Scholz & Schwarzer, 2006), and a review of 21 experiments found a small effect, φ = .24 (Carraro & Gaudreau, 2013).
Write down each day you did it
Across 138 randomised trials with 19,951 people, keeping track of progress helped people reach their goals, more so when they recorded it physically and when they reported it to someone (Harkin and colleagues, 2016). Telling ChatGPT how it went is a record too, though whether a chatbot counts as someone has not been tested.
When you miss
Tell it, and ask what got in the way. Put the question to the plan, not to yourself: in a large review of feedback, more than a third of it made performance worse, and the harm sat with feedback that turned attention to the person rather than the task (Kluger & DeNisi, 1996). Applying that to what you tell yourself is our reading, not the review’s. More in what to do the day after you miss a day.
If you’d rather it were built in
Everything above works in any chat, and it is free to start. If you’d rather have it built in, that is what Forvel is for: one thing a week on a card in your own words, a ring to log each day you do it, and a coach that, once you agree, reads how your last two weeks went.
Sources
- Sheeran, Listrom & Gollwitzer (2025). European Review of Social Psychology, 36(1), 162–194.
- Wood & Neal (2007). Psychological Review, 114(4), 843–863.
- Sniehotta, Scholz & Schwarzer (2006). British Journal of Health Psychology, 11, 23–37.
- Carraro & Gaudreau (2013). Psychology of Sport and Exercise, 14, 228–248.
- Harkin, Webb, Chang, Prestwich, Conner, Kellar, Benn & Sheeran (2016). Psychological Bulletin, 142(2), 198–229.
- Kluger & DeNisi (1996). Psychological Bulletin, 119(2), 254–284. Forvel has read this through summaries, not the paper.