About AI Training Pay
An independent pay index and reference for people who do AI training work.
AI training work pays anywhere from $10 to over $200 an hour, but the information about it is scattered across platform dashboards, Reddit threads and recruiter messages, some of them fake. AI Training Pay puts the public facts in one place: 1,094 live roles with pay converted to hourly, profiles of 15 platforms, and guides that cite their sources.
AI Training Pay is built and run by Hokuto Labs, an independent software studio. We are not owned by, and do not work for, any AI training platform or AI lab.
Editorial standards
- Data comes from the platforms' own public listings. We collect them automatically, convert pay to hourly only when the listing gives a real hourly figure, and show when each role was last seen. Full methodology.
- Every platform claim links to a source. Pay schedules, rates and policies are cited to the platform's own docs where they exist. When we rely on a third party, we say so. When nobody publishes it, we say that instead of guessing.
- Allegations are labeled as allegations. Lawsuits and worker complaints are described as claims, with their source, not as findings.
- Numbers are computed, not typed. Pay figures in our guides are calculated from live listings each time the site rebuilds, so they don't go stale.
- Guides show when they were last checked. We re-check platform facts when platforms change their terms, and update the date when we do.
How we make money
Some platforms pay a referral bonus when someone we send them starts working. Companies can also pay to feature a project, always labeled "Featured · paid placement". Neither changes how roles are ranked, what pay we show, or what we write about a platform. Workers never pay us anything. Full disclosure.
Corrections
If something here is wrong or out of date, tell us and we'll fix it. A corrections inbox is being set up; until it's live, we re-check platform facts whenever a platform changes its terms. Platforms can request corrections the same way; we'll update facts we can verify, but we don't remove accurate information on request.