A public, evidence-graded compilation of advertised hourly pay for people who do AI training work across platforms. It lists 14 platforms and 27 advertised pay bands ranging from $3 to $250 per hour, marking single-source claims separately and emphasizing that these are advertised rates, not guaranteed wages. Every band carries an evidence class (first-party listing, two agreeing guides, or single-source claim), sources are named and dated on the methods page, and a corrections log records verified changes. Values were transcribed by hand; the project was prepared by one person (Sept 2026) and invites corrections. Data and visual plates are available for download as CSV/JSON.
Beyond raw rates, the project answers five practical questions: pay, payout timing, access rules, quality ratings and ownership. It reports that 11 of 14 platforms publish payout schedules, there are three access tiers and six leading countries, four platforms publish user ratings (the lowest shown is 1.2 from 293 reviews), and Meta holds 49% market share. A “credential dandelion” visualizes pay bands on a log scale - highlighting a $20 crowd ceiling shared by Appen and TELUS and Mercor’s $250 top band - and cross-cut analyses cover timing, geography, sentiment and gatekeeping. Plotting code was drafted with an AI assistant but every figure is traced to source data.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.