What AI expert work pays
Every rate below is read from a live listing we track — no surveys, no self-reported figures, no projections. The figures change as listings open and close.
885 priced listings · 1036 tracked · computed 22 October 2026
885 listings publish compensation. 838 of them are hourly with a positive rate band, and every median and range on this page is drawn from them.
What the numbers say
Recalculated every time this page loads, from the 1036 listings currently open. Nothing below is typed by hand.
01
LAW pays the most, not GENERALIST
law listings run a median of $117/hr, topping out at $4050. generalist is the biggest category by volume — 138 of 838 rate observations — and sits at $70.
See law listings →02
The spread inside a field is wider than the gap between fields
law runs $31 to $4050 — a 131× range within one field, against a $30–$117 spread across all of them. The field alone does not settle which band a listing sits in.
Sort roles by rate →03
85% publish a rate before you apply
885 of 1036 listings state a band up front. 47 pay per task or per project rather than hourly, and are excluded from every median here rather than converted into one. The rest publish no figure, and no figure is estimated for them.
Browse every listing →Every field
| Field | Low | Median | High | Rate observations |
|---|---|---|---|---|
| LAW | $31 | $117 | $4050 | 80 |
| FINANCE | $30 | $100 | $350 | 82 |
| MEDICINE | $20 | $98 | $450 | 99 |
| ENGINEERING | $9 | $88 | $153 | 40 |
| SCIENCES | $30 | $85 | $400 | 101 |
| STEM AND CODING | $15 | $80 | $350 | 104 |
| OTHER | $20 | $73 | $150 | 8 |
| GENERALIST | $6 | $70 | $350 | 138 |
| CREATIVE | $5 | $60 | $140 | 45 |
| EDUCATION | $15 | $58 | $140 | 15 |
| LANGUAGES | $8 | $30 | $200 | 126 |
By platform
| Platform | Low | Median | High | Rate observations |
|---|---|---|---|---|
| AfterQuery | $15 | $80 | $175 | 228 |
| Ethos | $70 | $80 | $150 | 100 |
| Mercor | $10 | $79 | $450 | 232 |
| micro1 | $6 | $73 | $4050 | 258 |
How to read this
- Figures are the published bands on listings we track. They are what the platform advertises, not what every contributor is paid.
- Each listing counts as the midpoint of its advertised band. The median is the middle of those midpoints, so one unusually rich posting cannot drag it.
- Listings without a published rate are excluded rather than guessed at — 151 of 1036 right now.
- A field appears here once it has 3 or more rate observations. Below that there is a number but not a statistic, so we show nothing rather than a median standing on one posting.
- One listing advertises a floor of $0. It appears on its listing exactly as published, and is left out of the figures above — a zero floor is not a low rate, it is the absence of one.
- Rates are set by the platforms. We have no influence over them and earn nothing extra when they rise.
- For the same figures with percentiles, listing counts by field and platform, and the full method in one place, see the AI training market report.
The listings behind the numbers
Every figure on this page comes from a live listing you can open, with its published band, who it is open to, and what it screens for. Search by skill, language or credential.
Browse roles →What the screening asks
Mercor and micro1 both screen with an AI interviewer before a human sees you, and Ethos may invite you to an AI voice screen once you have created a profile. Each listing carries prep questions with what the question is testing and where candidates lose points.
Read the screening guide →Run the numbers on your own field.
Start from the median above, then set the rate you would actually accept. Nothing here is an offer — it is arithmetic on the rates the platforms publish.
Median on our listings for generalist is $70, across 138 rate observations spanning $6–$350.
An estimate, not an offer. What you are paid depends on your credentials and how you score in screening, and hours are never guaranteed — most people treat this as a supplement rather than a salary.