AI training work — labelling data, writing prompts, evaluating model outputs, annotating video — is contract work posted on a handful of platforms. The rates vary widely. Across the 597 open roles this site tracks, listed hourly pay spans $2–$400 an hour, and that span is real: the floor is a large-volume data labelling task and the ceiling is a niche legal or medical review.
Every figure comes from the platform’s own listing. We never estimate a rate, and we never convert a per-session or per-task fee into an hourly one.
Pay by platform
Four platforms supply the roles on this site today. Each sets its own rates, and the ranges barely overlap at the extremes.
micro1 lists 234 open roles. Hourly rates span $6–$400, with most falling between $15 and $75. micro1 posts the widest range because it covers everything from bulk data annotation to specialist expert review. It also accounts for the largest share of openings: over 59,000 of the 74,000+ stated across all platforms.
Mercor lists 260 open roles. Hourly rates span roughly $13–$270. Mercor skews toward mid-to-senior professional work — software engineers, data scientists, subject-matter experts — and its floor is higher as a result.
Terac lists 52 open roles. Most pay a fixed amount per session or per task rather than an hourly rate. Terac’s work often takes the form of short interviews, surveys or one-off evaluation tasks, so a per-session fee is the natural unit.
AfterQuery lists 51 open roles in a narrower pay band.
Hourly, per-session and per-task
Not every role pays by the hour. Of the 597 open roles:
- 540 list an hourly rate
- 40 pay a fixed amount per session (a single interview or evaluation)
- 17 pay per completed task
A per-session fee is not an hourly rate in disguise. When a platform says “$259 per session” and gives no duration, converting that to an hourly figure means inventing a number.
What drives the range
A $2-to-$400 span looks suspicious until you see what is at each end.
At the low end are high-volume, low-expertise tasks: data labelling, basic text classification, simple annotation work where the platform is hiring thousands of people and the task takes seconds each. The rate per task is low; the volume is high.
At the high end are roles that require deep domain expertise: a lawyer reviewing contract clauses for an AI model, a doctor evaluating medical diagnoses, a physicist checking mathematical proofs. These roles need credentials the platform cannot easily source.
In between, the largest cluster of roles pays $15–$75 an hour. This is where most generalist and specialist AI training work sits: prompt writing, response evaluation, coding tasks, content review.
What this does not tell you
These are listed rates. What you actually earn depends on availability (some roles have work for four hours a week; others have forty), how quickly you pass screening, and whether the role stays open long enough for you to be matched.
Actual hours worked, payout timing and task availability are things we do not yet know enough about to state.
Every figure on this page updates with each morning’s check. Browse the full list of open roles to see what is available now, or compare platforms on the pay rates table.