Research access to the AI-training pay dataset
HumanSourcer computes a live, evidence-cited pay index for AI-training and data-labeling gig work - the same figures published on the Pay Index and Market Data pages. If you're studying the gig economy, AI-labor markets, or platform pay dynamics, you can request a structured export of the underlying dataset rather than re-deriving it from the rendered pages.
The median advertised rate for AI-training work is $50/hr (HumanSourcer Pay Index, September 2026; n=700 distinct advertised rates across 74 platforms). Full methodology is documented at how this figure is derived.
What's in the export
Pay figures parsed from live listings, deduplicated by platform and advertised rate so a single repeated posting doesn't vote more than once - the same discipline behind the published median.
Each rate tagged by work domain (coding, linguistics, science/STEM, etc.) and task type, matching the taxonomy the site's own hub pages use.
Whether a rate came from a structured field, was parsed by rule, or was extracted from listing prose by a model - and whether the source text hedged the number.
Which of the 74 tracked networks each rate belongs to, with public ownership/relationship data already sourced on each provider's page.
Embed the live figure
If you cite the Pay Index, this badge keeps the number current instead of freezing it at the date you copied it - it re-renders daily from the same data as the figure above.
Paste this on your own site:
<a href="https://www.humansourcer.com/ai-training-pay/"><img src="https://www.humansourcer.com/og/badge.png" alt="HumanSourcer Pay Index - median advertised AI-training pay rate" width="340" height="68" loading="lazy" /></a>
Request access
A sentence on what you're studying helps get your request answered faster, but isn't required.
Frequently asked questions
What data is actually in the export?
The same derived dataset already published in aggregate on the Pay Index and Market Data pages: normalized, deduplicated hourly pay rates by platform, domain, and task type, with the parsing/estimate provenance for each figure. It is the computed layer, not raw scraped listing text.
Is this a live API?
No - access is a structured export (CSV/JSON) granted by hand after a request, not a self-serve API key. That's a deliberate choice: it keeps this offer real today without reopening how HumanSourcer handles bulk automated access to the underlying data.
Is it free?
Free for academic and non-commercial research use, with attribution to HumanSourcer and a link back to the Pay Index. Commercial use is a separate conversation - say so in the request.
How long until I hear back?
There's no fixed turnaround - requests are reviewed by hand, not queued through an automated system. Describing what you're studying in the request helps it get answered faster.