Press and media

Data on what AI training work pays, built from the platforms' own public listings. Every figure below comes from a frozen monthly report, so it won't change after you cite it.

The next report publishes November 1, 2026

Quotable findings appear here when it does. Until then, the live Pay Index and pay trends show today's figures, and the October 2026 edition is archived with its restatement.

Fact sheet

What it is: An independent index of AI training jobs and pay, with every rate converted to hourly where the listing allows it.
Publisher: Hokuto Labs. Founded and edited by Michael Cortez.
Data: Every live listing on the public job boards of Mercor, micro1, Handshake AI, Alignerr and OpenTrain AI: 1,785 roles today, refreshed daily, with daily history since October 19, 2026.
Not in the pay figures: Outlier, DataAnnotation, Prolific, Invisible, Turing, Remotasks, Appen, TELUS Digital AI Community, Toloka and Braintrust have profiles on the site but aren't in the pay figures, which come only from the 5 job boards above.
Report: The State of AI Training Pay is frozen on the 1st of each month and never revised after publishing. The next edition is due November 1, 2026.
License: Text, charts and data are free to reuse under CC BY 4.0 with credit to AI Training Pay and a link.
Independence: No platform pays to change its ranking or what we write. Some platforms pay referral bonuses, and companies can buy clearly labeled featured placements: full disclosure.

Who to talk to

Michael Cortez

Founder and editor, AI Training Pay

Founder and editor of AI Training Pay. Search and AI strategist with nearly 20 years in digital marketing. He can walk through the method, the data behind any figure, and what it can't show.

What the data can and can't show

These are listed rates: what platforms advertise for a role. They aren't earnings, and platforms don't guarantee hours.
Each role counts once, at the midpoint of its hourly range. Per-task pay is left out, and so are roles that list only a ceiling (“up to $X/hr”), because neither has an honest hourly midpoint.
Month-to-month changes compare only the platforms covered in both months, so adding a platform never reads as a change in pay.
Coverage is the platforms with a readable public job board. It isn't the whole market.

Full methodology →