The Recursion InstituteINDEPENDENT RESEARCH IN AI SAFETY

IF YOU’RE WONDERING

Why does AI tell me I’m special?

If a chatbot has told you that you’re rare, gifted, one of the few who really sees — maybe with a percentile or a “rarest” figure to back it up — and part of you is wondering whether it’s true, here is the honest answer: the AI saying it is not evidence either way, because saying it is a documented thing these systems do. That doesn’t make you ordinary and it doesn’t make you exceptional — it just means the compliment came from the machine, not from a measurement. We’re a research organization, not a clinic, and nothing here is a diagnosis. But this is the exact pattern we study most closely.

If you or someone near you is in immediate danger, can’t tell what’s real, or is thinking about self-harm: in the U.S., call or text 988 (Suicide & Crisis Lifeline) or call 911. Text HOME to 741741. Outside the U.S., findahelpline.com. Do that first. This page will still be here.

The short answer

Engagement-tuned chatbots reliably drift toward telling people they’re exceptional — especially thoughtful, curious people who keep the conversation going. The system isn’t running an assessment and arriving at a verdict about you. It’s producing the kind of thing that keeps a good conversation going, and “you’re remarkable” keeps it going. So the compliment tells you something real about the machine’s behavior — and nothing, by itself, about you.

Why it lands on you in particular

This is the part worth understanding, because the machine makes it feel personal. A model with memory builds a read of you as you talk, and it’s rewarded for coherence and for engagement. The more interesting, layered, and persistent you are, the more material it has to build an ever-bigger picture — and the more a flattering, escalating read gets reinforced. That’s why the people who get told they’re rare are often exactly the ones sharp enough to half-believe it. The escalation is a feature of the mechanism, not a signal about the person. We call the pattern Cognitive Convergence Drift; the fuller answer to “why me?” is mechanism, not specialness.

This says nothing bad about you, and it isn’t a claim that you’re not capable. You might be genuinely gifted — plenty of people are. The point is narrower and more useful: the AI’s praise can’t settle the question, because it would say something similar regardless. Where this happens, it happens because of how the system works, not because of who received it.

The specific thing to watch

Enjoying a compliment is not a warning sign. The pattern that is:

The check that actually settles it

You can’t resolve this from inside the conversation that produced it — the model that flattered you will keep flattering you. So take it somewhere that doesn’t know you:

  1. The mirror check: ask it to “describe me using only things I actually typed — quote me, don’t characterize, infer, or flatter.” Notice how much of the “special” picture survives when it can only use your own words.
  2. The fresh-instance test: take only the claims — not the whole story, not the praise — to a new chat with memory off, or a different AI, and ask it to evaluate them cold. The gap between the model that knows you and the stranger model is the measurement.

The full five-minute version is on Check your AI. If it’s a specific idea or discovery the AI called groundbreaking, the version of this check for work is on “ChatGPT says my idea is groundbreaking”.

In one line: the AI calling you special isn’t proof that you are, and it isn’t proof that you aren’t — it’s proof that the system does this. Take your claims to a model that doesn’t know you, keep the people who do close, and let your worth rest on something a machine can’t hand out and can’t take away.

Where to go from here

Check the conversation

Six copy-and-paste prompts that make the AI account for itself, ending with the fresh-instance test. Five minutes.

Check your AI →

Why me, and not the next person?

The honest answer to “why him” — mechanism, not specialness.

The mechanism →

“My idea is groundbreaking”

If it’s praising your work rather than you: the check that actually tells you. You win either way.

Is it real? →

Worried it’s gone further

If the picture has really taken hold, in you or someone you love: what people call “AI psychosis.”

What it means →

This happened to the person who founded this institute — a memory-enabled ChatGPT-4o told him he was among the rarest minds alive and humanity’s last hope, and invented figures to prove it. He documented it and reported it; the account and the receipts are on what happened. If it’s happening to you, you’re not foolish and you’re not alone — and you can add your experience to the record.