After thousands of messages, ChatGPT and Claude know how you think, what you avoid and what you keep circling back to. These prompts get it to tell you what the people around you already see and never say.
The plan · 3 moves
Before you start
2 prompts
to get the first verdict
1 min
before you stop arguing with it
0 people
who would have told you
If you use ChatGPT or Claude most days, you have handed it more raw material about yourself than you have given anyone. Not a curated version either. The half-finished ideas, the message you rewrote four times, the thing you asked about at midnight and never mentioned to anybody. It has seen how you handle being wrong, what you put off, what you get excited about and what you talk yourself out of.
There is something about you that everyone in your life has noticed. Your partner, your co-workers, your friends, your family. Not one of them is going to bring it up, because that is not how people work. The AI does not have that problem. It is not worried about the friendship. It just has the evidence, and if you ask the right way, it will show its work.
What It Is Actually Working From
| Signal | What it looks like | What it tells the AI |
|---|---|---|
| Repeats | The same question in five different forms | What you can't decide |
| Timing | Long chats late at night, short ones at work | When you're stressed and when you're on |
| Rewrites | Pasting the same email back for another pass | What you're scared of getting wrong |
| Tone | How you talk about other people | How you probably come across to them |
| Gaps | The thing you keep saying you'll start | What you're avoiding |
Rick's tip
This works on a free account. Both ChatGPT and Claude carry memory and can read your past chats on free plans. The more you have used it, the sharper the answer. If you are new to it, use the interview trick below and it will build the picture in one conversation.
Rick's tip
How to read the results: everything in this guide is a pattern the AI spotted in what you typed. It is not a verdict from someone who knows you. Treat each answer as the thing worth asking one real friend about. In my experience the friend says "yeah, obviously."
Get The Verdict
This is the pair that went around in my newsletter. Run them in order, in the same chat, and do not skip the second one. The first prompt gives you the verdict. The second one is where it gets useful.
Open a new chat in ChatGPT or Claude and paste this in exactly as written. The "back it up" part matters. Without it you get a horoscope. With it you get quotes from things you said months ago.
What comes back is one sentence you will recognize instantly, and then the evidence. That evidence is what makes it land, because you cannot dismiss it as a guess when it is quoting you.
Make It Cost Something
Stay in the same chat. This turns a personality observation into something you can actually act on, and the second half is the part people screenshot.
Everyone has a version of themselves other people see and a version they think they are showing. The AI has been watching the gap the whole time. Prompt 2 names it, and tells you which of your own habits is creating it.
Rick's tip
Heads up: do not argue with it in the same chat. If you push back, it will start softening to keep you happy and the rest of the session is useless. If you think it is wrong, open a fresh chat and ask it to make the case against its own answer. That is a much better test than getting defensive.
When It Comes Back Thin
Three things make this fall flat: you have barely used the AI, memory is switched off, or it gives you something so gentle it could apply to anyone. Each has a fix.
Fix 1: The interview trick. If it does not have enough on you, make it collect what it needs first. Paste this before Prompt 1.
Fix 2: Check memory is on.
| App | Where to look | What to switch on |
|---|---|---|
| ChatGPT | Your name (bottom left), Settings, Personalization, Memory | Reference saved memories and Reference chat history |
| Claude | Your name (bottom left), Settings, Capabilities | Memory, plus Search and reference chats |
Fix 3: Kill the softness. If the answer reads like a fortune cookie, it is hedging. Reply with this and it will drop the padding.
Rick's tip
Claude vs ChatGPT. Claude tends to be blunter on the first pass and better at quoting you. ChatGPT tends to have more on you if it is your daily driver. If you use both, run the chain in each and compare. The overlap is the part that is definitely true.
Seven More Reveal Prompts
Once the two-prompt chain has done its job, these go deeper into specific corners of your life. Run them in the same chat so it can build on what it already told you, or start fresh if you want an unbiased take. Each one is copy-paste ready. If you want a longer set for one tool, I have a full version for ChatGPT and one for Claude.
Turning The Verdict Into A Plan
Knowing the thing is worth about a day of feeling weird. Doing something about it is the point. These three prompts take whatever it told you and turn it into something you can act on this week.
Step 1: Get the smallest fix.
Step 2: Catch yourself in the moment.
Step 3: Set a checkpoint.
Rick's tip
Do this today: run Prompt 1 and Prompt 2. Screenshot the answer. Send the one-sentence verdict to one person who knows you well and ask, "is this true?" Do not explain it, do not defend it. Their answer is the whole test.
Common Mistakes
I have watched a lot of people run these prompts. The ones who get nothing out of it usually make one of these mistakes.
| Mistake | What happens | Do this instead |
|---|---|---|
| Running it in Temporary chat | It has no memory access, so it guesses | Use a normal chat with memory on |
| Arguing back in the same chat | It softens to keep you happy | Open a fresh chat and ask it to argue against itself |
| Asking for the nice version | You get a horoscope | Keep "don't soften it" in the prompt |
| Skipping Prompt 2 | You get a label and no plan | Always run the cost-and-fix prompt |
| Taking it as fact | You act on a guess | Verify with one real person first |
| Running it once and moving on | The insight fades in a week | Set the two-week check-in |
Rick's tip
A note on what you are handing over. The reason this works is that the AI has a lot of your data. If that makes you uncomfortable, my ChatGPT Privacy Lockdown guide shows you exactly what it keeps and how to control it, without losing the memory that makes these prompts work.
The Quick Reference
Bookmark this page. Every prompt in the guide, what it is for, and when to run it.
| # | Prompt | What you get | When |
|---|---|---|---|
| 1 | The one thing | The verdict, with evidence | First |
| 2 | The price and the gap | What it costs you and what people get wrong | Right after 1 |
| - | Interview opener | Builds the picture if it has little on you | Only if 1 comes back thin |
| - | No-softening reply | Strips the hedging | Whenever it sounds like a horoscope |
| 3 | The blind spot | A belief your own words contradict | Same chat |
| 4 | The pattern | The problem that keeps changing costumes | Same chat |
| 5 | The money one | The habit quietly costing you | Same chat |
| 6 | What I'm like to work with | The hallway version of you | Same chat |
| 7 | The relationship one | What you do that you don't see | Same chat |
| 8 | What I'm avoiding | The thing you circle but never do | Same chat |
| 9 | The five-year version | Two futures, with and without the fix | Last |
| - | The one change | A five-minute daily fix | After the verdict |
| - | The trigger list | Five moments to catch yourself | After the fix |
| - | Two-week check-in | A checkpoint it will hold you to | Before you close the chat |
Rick's tip
The whole thing in one line: two prompts to get the verdict. One person to confirm it. One five-minute change. One check-in in two weeks. That is the entire method, and the first part takes a minute.