Blind tests. Verified humans. Real answers.

Put two (or more) options in front of a crowd where every vote is a unique, verified human — and find out which one actually wins.

Audio, video, images, or text. Labels hidden until after the verdict. One human, one vote — enforced by World ID proof-of-personhood, not by promises.

Take a live test →

Media quality — which is best? head-to-head Rate each voice (1–5) star rating Rank the encodings ranked choice Clarity knockout — best of four tournament Which will most people prefer? majority predict Quick picks — go with your gut rapid swipe

Just looking? Try the preview sandbox — no sign-in, votes don't count.

How it works

Verify once

Prove you're a unique human with World ID — without revealing who you are. Works on the web and inside World App.

Judge blind

Experience each option — listen, watch, read — then pick the best. Names stay hidden until after you choose; order is shuffled every time.

See the verdict

Labels unmask, results update live, and on majority-reward tests you earn points when your verdict lands with the crowd.

Why the results are trustworthy

One human, one vote

Every verdict is tied to a World ID proof-of-personhood. Bots, scripts, and ballot-stuffing simply don't count.

Blind by construction

The system behind each option never reaches your browser — not in the page, not in the network. Position is randomized per person.

Attention enforced

A vote only counts if you actually experienced every option — play time, reading time, and decision time are all checked server-side.

Precognition research

The same verified crowd powers an open research programme: can anyone predict which picture a public randomness beacon is about to choose? Every target is drawn from the beacon after the prediction is locked, and every trial can be rechecked afterwards by anyone from public data. Nothing here is settled — the protocols are published, fixed in advance, and designed so a negative result is just as reportable as a positive one.

Prediction Practice free · four pictures · scored against chance The timing study pilot · 5, 10 or 20 seconds to decide The research proposal protocol, live demo and references

Run your own test

Comparing voices, video cuts, thumbnails, or copy? IABT puts your options in front of a verified human crowd — blind, Sybil-proof, with results you can act on. Templates include head-to-head, star rating, ranked choice, tournament, majority-predict, and rapid-swipe.

Talk to us — contact@iabt.com