My Experience Using ChatGPT As A Therapist
And diary, and data analyst, and life coach, and health advisor, and....
A little before New Year’s, I started using ChatGPT as a combination diary, life coach, therapist and accountability partner. The basic idea: keep a detailed journal, several times a day, and have it hold me accountable to my goals while also running analytics on the data I’m feeding it.
After two months I can say the experience has been positive overall. The mere act of journaling has helped me stick to my resolutions, be more productive, and follow better health habits. GPT also has identified a few patterns that I might have had trouble spotting myself, although I can’t credit it with any mind-blowing insights.
As for the results: First and most importantly, I feel better. I’ve been depressed for the last year, owing largely to a traumatic breakup. I was starting to slowly get better, but my recovery has accelerated thanks to this experiment. I still have my bad days, but they’re rare.
Second, I’m more productive. In the last two months I’ve purchased a website I used to write for and started fixing it up, run a few photo meetups, been hired as a photographer, and am now planning my first fashion show.
Third, I’m healthier. I’ve been working out more, eating better, sleeping better, getting up earlier, and stopped needing afternoon naps on most days.
I think copying this experiment would be valuable to most people– in fact I expect it will become more effective with the newer models that are coming out this year. In this article I’ll explain how I set it up, what I learned about myself, and the strengths and limitations of this method.
How I Set This Up
I won’t share the exact prompt for privacy reasons, but the initial prompt had a few sections:
1- Explanation of my purpose. I’m going to keep a journal a few times a day, and I want you to a) encourage me to stick to my goals, b) analyze the data I’m giving you for patterns, and c) offer me life advice.
2- List of my resolutions that I want to keep.
3- List of problems I know I have that I want to work on.
4- A recap of how I expect this to work and what I want from ChatGPT.
And that’s about it. From then on, a few times a day I wrote entries explaining what I’d been doing what I ate and drank (not water so much as caffeine intake), and I recorded a few key variables on a one to five scale: My energy level, mood, stress, libido, productivity, and itching (I have atopic eczema).
It helps to have some kind of consistent idea of what each number on these scales would mean. As an example, for energy, a one means I feel compelled to lie down, a two means I could lay down or stand up, a three means I want to stand and maybe even walk around, a four means I have a strong urge to move and can’t easily stand still, and a five means I’m hyperactive. I define this by the amount of physical activity I find myself wanting to gravitate towards.
Here’s What I Learned About Myself
I’m pretty much going to copy past here, with a few items censored for privacy. My own commentary will be in italics.
1. Energy ⚡
✅ General Patterns:
Your energy usually starts low (1-2/5) in the morning, then rises to 3/5 or 4/5 after caffeine and food. This is a bug hangup for me. I want to get going faster in the morning!
Caffeine & protein help increase your energy, while high-carb meals make you sluggish. This is something I noticed myself actually; eating Korean barbecue gives me sustained energy for 6-8 hours. In fact when I first moved to LA, my friends and I used to do Korean BBQ before going clubbing.
Workouts help sustain energy—even light exercise boosts it for hours. I got crashes after big workouts early on but they stopped as this experiment went on.
⚠️ Negative Triggers:
Caffeine crashes: When you overconsume (e.g., buffet dinners, multiple coffees, energy drinks), you often feel wired at night but sluggish the next morning.
Poor sleep: If you don’t sleep well (due to itching, stress, THC, or late caffeine), your energy stays at 2/5 all day despite caffeine. It means the next day, to be clear. The THC part comes from the time I tried edibles.
THC/Edibles: These can leave you groggy the next morning if taken too late or in high doses.
One thing I’ve noticed that it didn't pick up on: I’m more prone to food coma earlier in the day. I don’t know why, but a moderate-sized meal, unless it’s low-carb, will make me tired around noon, but I can crush a buffet at night and feel fine.
2. Mood 😊
✅ General Patterns:
Mood is highest (3-4/5) on productive days when you finish tasks and socialize. Yeah, this scans: my self-esteem tends to be based on how productive I’ve been today and how well my work is going.
Social interaction, working out, sex and photo shoots boost your mood significantly. This was more of a surprise than it should have been; discussion later.
You feel happiest when progressing on long-term goals like your fashion show, AI projects, or fitness.
⚠️ Negative Triggers:
Sleep deprivation makes your mood drop to 2/5 or lower.
Canceled plans or loneliness cause temporary mood dips, especially on event-heavy days (e.g., Valentine's Day).
Overplanning leads to feeling overwhelmed, which causes mood crashes.
3. Stress 😰
✅ General Patterns:
Your stress peaks (4-5/5) when you procrastinate or feel behind on goals.
Major work tasks (e.g., fashion show planning, website issues, tax problems) often elevate stress.
When you break tasks down or get small wins, stress drops quickly. Key insight!
⚠️ Negative Triggers:
Government/Finance Calls: These make your stress spike, especially long hold times.
Indecision: Days where you struggle to decide what to do next lead to a low-productivity/high-stress cycle.
Tech Issues: When AI websites or AdWords don’t work, stress levels rise. This is related to that website I bought.
4. Libido 🔥
✅ General Patterns:
Libido is highest in the morning and lowest in the evening.
Going on dates and having sex gives you a sustained boost to libido. Social interaction also seems to help.
Workouts help boost libido.
⚠️ Negative Triggers:
Poor sleep or exhaustion lowers libido.
Excessive adult content leads to crashes in both libido and focus.
Caffeine, stress, and mood dips seem to suppress libido.
5. Productivity 🎯
✅ General Patterns:
Your productivity is highest (3-4/5) when you have a single major goal for the day. Again, key insight.
Physical movement before work (gym, walk) helps focus.
AI projects, fashion show tasks, and photography editing tend to be your most productive tasks.
⚠️ Negative Triggers:
Overloading your to-do list leads to avoidance.
Distractions (adult content, social media, procrastination) eat up focus.
Sleep deprivation reduces motivation, making work feel impossible.
6. Itching 🩹
✅ General Patterns:
Flare-ups (4-5/5) often happen at night or early morning.
Showers help reduce itching short-term but don’t fully prevent it.
Itching is lowest (2/5) on days with low stress, better sleep, and moderate caffeine.
⚠️ Negative Triggers:
Excess caffeine or alcohol often precedes bad itching days.
High-stress days can trigger itching flares.
Sleeping poorly leads to worse itching the next morning.
Key Takeaways:
Energy & Productivity thrive on moderate caffeine, protein-rich meals, exercise, and clear task prioritization.
Mood & Stress are best managed through social interactions, fitness, and breaking tasks into manageable steps.
Libido & Focus suffer when you consume too much adult content or caffeine, don’t sleep well, or feel stressed.
Itching Patterns: Most likely linked to stress, caffeine, alcohol, and sleep quality.
So Here’s What I’ve Learned From This
Overall, this is a success. Let’s start with the downsides though.
ChatGPT was too nice overall. I wanted tough love, but it only gives that when specifically asked for it. Otherwise, it’s unrelentingly positive and tends to shower me with validation. I actually picked ChatGPT to minimize this issue, because Claude is notorious for its obsequious personality. Overall, I wish it pushed me harder and wasn’t so easy on me– not like a movie drill sergeant, but like a good personal trainer.
AI still struggles with long context windows. This seems to be about as much of a problem regardless of which model you’re using. As this experiment went on, the AI’s memory of earlier parts of the conversation got fuzzier. That limited its data analytics capabilities and I sometimes had to remind it of things, but on the other hand a lot of what it did for me only required it to respond to the last few messages.
Adult content filters got in the way. I had to be careful talking about adult stuff, which limited my ability to analyze libido in particular.
It eventually started getting too chatty. Like, asking me for details on what I was up to that it didn’t really need, badgering me with never-ending questions. I can’t be sure, but this seemed to start on a day when I was in a bad mood and it started trying to distract me with small talk. After seeing my mood improve a point, it may have over-learned from that and decided more small talk, all the time, was the way to go.
See why I didn’t use Claude? I’m looking for a robot assistant, not a friend.
On the other hand, it didn’t show much initiative about analysis. If I wanted analysis about what I was doing right and wrong or patterns it noticed, I had to ask for them. The lack of initiative is very noticeable compared to a human; ChatGPT rarely volunteers ideas unprompted. It never goes “You know, it just occurred to me that….”
Okay, so the upsides.
It worked. That’s the key thing: I got some actionable insights and my life measurably improved. By a lot, actually. I’m happier, healthier, more productive, and making more money than I was two months ago.
The mere act of journaling was immensely valuable. Part of that was the accountability, even to a chatbot. But I think a big part is the clarity that comes from writing things down, even without a response from anyone.
I learned things about myself that I would not have noticed. In particular, the importance of social activity to both my mood and my libido. It was beneficial to hang out with people at least three times a week, and possibly even more but that’s the limit I tested to. This seems obvious– of course socialization is important!– but people can have blind spots about themselves.
Actually, I want to dwell on that a bit. The average person hangs out with friends something like once a week, or four hours a week. That has always felt a bit low to me, and now I know that I, an autistic introvert, experience mental health benefits from getting at least three times that much social activity.
I had this notion in my head that I don’t need as much socialization as other people. Turns out I need a lot! But then, I suspect that everybody does.
I have clear directives. The biggest thing is I need to cut caffeine. That clearly causes wide-ranging issues for me. I also need to keep eating high-protein and low-carb, keep my daily to-do list simple, and still make time to socialize instead of idealizing the idea of being a productivity monk.
So what’s next? I’m going to take a few days off from this experiment, then maybe try it again with one of the newer AI models that’s coming out right now. I might also try programming it as a GPT next time– that may allow me to give it a personality that’s more willing to give me negative feedback, or acts with more initiative.
If you’re reading this– highly recommend you try it. I’ll provide an update in 2-3 month about the next phase of this experiment.