The box nobody ticks
YouTube gives creators a checkbox that says 'this video contains paid promotion'. Across 227,632 sponsored videos, we counted how often it is used. For commission deals: less than one time in twelve.
6 min readResearch · 6 posts · what we measured, in plain English
Method · 7 September 2026We test our own metrics.
And publish it when they’re wrong.
We shuffled 30,000 creators at random
Rival advertisers are supposed to hire different kinds of creator. We checked 2,725 of them, then dealt the same creators back out at random. Between 65% and 93% of the difference between advertisers turned out to be luck.8 min readof the difference between advertisers was luck65–93%depending on the measure, from the postDoes a brand book you twice?
Half of real paid creator relationships never happen again. We counted 108,433 of them — and found that the bigger the creator, the only slightly better the odds.
7 min readThe fake-follower test that points backwards
Almost everyone uses views-per-subscriber to spot a bought audience. We checked it against 1,310 channels YouTube actually cleaned up. It does worse than guessing — and there is a simple reason why.
7 min readWhat actually predicts a fake audience
Six popular ways to spot a fake audience, put through the same test. Three point the wrong way, one has quietly stopped working, and one is genuinely good.
6 min readSmall creators don't engage better
Brands pay a premium for small creators because their audiences supposedly react more. Across 143,000 channels, mostly they do not — and the exceptions depend entirely on the subject.
6 min readEvery post links the data it was measured on.