Prove It or Park It

Most “is this AI?” drama is the wrong question. It doesn’t matter who typed the keys if the claim can’t survive daylight. We treat detection like a party trick and ignore the thing that would actually make us safer: provenance.

If a paragraph matters—if money, health, or reputations hinge on it—ask for a receipt. Not a vibe. A chain you can follow without becoming a monk. One link is plenty. If the link is missing, the sentence can stay; it just can’t steer.

I use a tiny audit loop now. One: mark the sentence with teeth. Two: can I verify it in a minute? Three: if yes, capture the source so Future Me doesn’t repeat this chore. Four: if no, label it “color” and keep it away from decisions.

This isn’t about killing fun. It’s about preventing your brain from outsourcing judgment to a confident paragraph. You wouldn’t accept a GPS that never shows the road, only soft jazz and a thumbs-up. Don’t accept it in writing either.

If you’re tired of arguing about detectors, there’s a short guide that reframes the game: Is this AI-generated? Stop guessing, start testing. And if you want a neutral grown-up rulebook, NIST’s AI 100-4 synthetic content guidance is where standards bodies are pointing. Neither one asks you to turn into a hall monitor. Both ask you to separate facts from furniture.

The best part of requiring receipts is not moral purity. It’s speed. Once you practice, you know in thirty seconds whether a paragraph can be used for anything important. The rest is just decoration.

So, prove it or park it. The internet doesn’t need more detectives. It needs more people who can calmly say, “this part holds; that part doesn’t,” and move on with their lives.