Most AI detectors hand you a confident percentage - even when the evidence doesn't support one. AI or Authentic does the opposite: it layers C2PA provenance, SynthID watermarks and its own video model, and when the signals contradict it says "not decidable" instead of inventing a number. Paste a link (TikTok, Instagram, YouTube) or upload a file. You get a calibrated result, a plain-language breakdown of what each check found, and a shareable certificate. Free, no signup.
No reviews yetBe the first to leave a review for AI or Authentic?
Maker
📌
Hi Product Hunt 👋
I built this after a relative shared an AI-generated "news" clip in our
family chat. Nobody could tell it was fake. So I ran it through the
existing detectors - and every single one handed back a confident
number, on footage where they were obviously guessing.
That bothered me more than the fake did. A detector that's confidently
wrong is worse than no detector at all.
So AI or Authentic is built abstention-first. It checks C2PA
provenance, SynthID watermarks, deepfake face signals and its own
video model - and when those contradict each other, it tells you
"not decidable" instead of inventing a percentage. Every result
explains what each check actually found, and comes with a shareable
certificate, because the point isn't being right alone. It's ending
the argument in the group chat.
Free, no account needed. The EU's AI labelling rules just kicked in,
so I suspect a lot more "wait, is this real?" moments are coming.
One ask: drop a link to a piece of media you genuinely couldn't call.
I'll run it and post what the check found - including the ones where
it honestly can't decide.