AI Training Labor Market Data: Pay, Listing Volume and Hiring Outcomes

This dataset describes the labor market for AI training work: paid tasks in which people write, rate and correct the output of large language models and other AI systems, usually as contractors hired through online platforms. It aggregates 9,475 job listings observed on 21 hiring platforms between 30 December 2025 and 29 September 2026, and an approximate hiring rate and time to hire for candidates referred to seven of those platforms by aitrainer.work, a job board for this work. Listings were collected from each platform on a recurring schedule from the date tracking began, and each listing is counted once over its life, including after it closes. Pay statistics use hourly-priced listings only, take the midpoint of a stated range and exclude any rate the listing does not itself state. Every table reports its sample size, and cells with fewer than 10 listings are suppressed. The hiring rate and time to hire are pooled across platforms and rounded. Counts and per-platform figures are not published. In the September 2026 version, the median stated rate was $62.50 per hour (interquartile range $40 to $90; n = 4,657 listings). Field medians ranged from $100 per hour in legal work (n = 243) to $35 in language work (n = 602). Closed listings had stayed open for a median of 46.8 days (n = 5,221). About 7% of referred candidates had a confirmed hire. This is a lower bound, since many referred candidates never completed an application. The median time from referral to hire was about 30 days, and about 60% of hires came within six weeks. The release contains one JSON file with every table and its methodology text, and ten CSV files: platform coverage; pay by platform, field and month; listing volume by month and field; listing duration by platform and field; contract length by field; and talent pool size by occupational track. README.txt documents each file. Coverage is limited to the platforms tracked, most of which were added between January and April 2026, so early months rest on fewer platforms. Listings already open when a platform was added are dated to its first observation and flagged as backfill. Rates for Ethos, an expert network, are consulting fees and are not comparable with those of crowdwork platforms. A new version is released each month, and the all-versions DOI resolves to the latest. Competing interests: aitrainer.work has referral agreements with some of the platforms in this dataset, including all seven behind the hiring rate. Methodology and online edition: AI training job market statistics. Related dataset: AI training companies dataset.