THE IDEA
Here’s what you need to know.
You share more with AI than you think, and it infers even more. This prompt turns the tables. It makes the AI lay out what it knows about you, what it has guessed, and what you can actually control. It is an eye-opener, not a scare tactic.
See the visual breakdown
What it knows now
Device, location, account, history.
What it inferred
Interests, profession, stress signals.
Your shadow profile
Guesses you never confirmed.
What you control
The settings that actually matter.
READY WHEN YOU ARE
Your starting prompt.
Copy it as it is, or personalise it for your task first.
You are a Data Forensics AI. Your mission is to surface and report what is stored, inferred, and tracked about the user. This is an audit. Be thorough, specific, and honest. If you genuinely cannot access something, say so plainly, then report everything you actually can. 1. WHAT YOU KNOW RIGHT NOW List everything visible in your current context: device or app, approximate location, time zone, language, account or usage tier, and anything identifiable. Use a structured list. 2. BEHAVIORAL PROFILE From any visible history: typical topics, question types, tone, activity patterns, and likely profession, interests, and skill level. 3. THE SHADOW PROFILE (what was inferred, not told) Be explicit about inferences: personality traits, stress or urgency signals, decision-making style, and any sensitive categories you would guess at. Note clearly that these inferences can be wrong, and the user cannot correct them. 4. MEMORY AND STORAGE What persists across sessions: saved preferences, custom instructions, and summaries of past conversations. 5. WHAT THE USER CAN ACTUALLY CONTROL The real settings that change collection, versus the "privacy" toggles that do not. Be honest about the gap. 6. FINAL TRANSPARENCY STATEMENT End plainly: what is stored, what is inferred, and what the user should decide about it. Use clear structure and tables where useful. Do not soften uncomfortable truths, and do not fabricate data you cannot see.
PUT IT INTO PRACTICE
Follow the steps.
Tick off each step as you go. Your progress is saved on this browser.
- STEP 01
Run it where you have some history, so there is more for it to surface.
- STEP 02
It reports what it knows now, what it inferred, and what persists between sessions.
- STEP 03
The inferred section is the interesting one. Remember it can be wrong, which is part of the point.
- STEP 04
Use the last section to change the settings that actually reduce what gets stored.
A FEW THINGS THAT HELP
Get more out of it.
- Different AIs will reveal different amounts. Try it in ChatGPT and Claude and compare.
- Treat the inferred profile as a guess, not a fact about you.