Original research · last measured 2026-08-16
Half the roles that publish their hours aren't full time
Every hourly rate you read for this kind of work gets multiplied by 40 hours somewhere along the way — in a salary tool, in a Reddit comment, on a job board. Of the 251 listings here that say how many hours they actually want, 118 ask for less. That is 47%, and it moves the annual figure by more than most people would guess.
What the 40-hour assumption costs
Across the 112 part-time roles that publish both a rate and their hours, quoting the rate as a full-time year overstates it by a median of $56k.
At the extreme it is worse. The largest gap in the catalogue is $299k — a 15-hour-a-week role advertising up to $230 an hour, which reads as $478k a year on a full-time assumption and $179k on its own published hours.
This site had the same bug until the day this page was written: five separate places annualised every hourly role at 40 hours. That is how the study started.
The hours actually published
| Hours / week | Listings | Median rate |
|---|---|---|
| 40 · full time | 133 | $75 |
| 20 | 84 | $36 |
| 35 | 12 | $75 |
| 15 | 11 | $170 |
| 10 | 11 | $78 |
The rate column looks like shorter weeks pay better. They don't — that is a field effect, and reading it the other way round is the mistake this table invites. The 15-hour band is pharma and clinical consulting: trial endpoints, market access, epidemiology, at $170 median. The 20-hour band, 84 listings and the largest part-time group, is mostly music, language and creative-review listings at $36. Specialists get hired for short weeks. Short weeks don't pay specialists' rates.
Method, and what this can't tell you
Only one platform in this directory publishes a weekly commitment at all. 251 roles carry the field and 513 do not. Every percentage on this page is taken over the 251 that publish — never over the 764 in the catalogue — because a listing that says nothing about hours is unknown, not full time. Treating silence as 40 hours is exactly the error the page is about.
Rates are the midpoint of the advertised band, hourly listings only. Per-task fees and annual salaries are excluded rather than converted, and the 40-hour band is the largest simply because full time is still the commonest arrangement — this is not a claim that most of this work is part time.
Where a platform publishes hours, this site now uses them. Where it publishes none, the role page still shows a full-time estimate and says so in words on the page, rather than presenting the assumption as a measurement.
Figures move with the catalogue, which is re-read daily. Last measured 2026-08-16. See also what the work pays by field and who is allowed to apply.
Common questions
Are remote AI training jobs full time?
Often not, and the listing usually doesn't lead with it. Of the 251 roles in this directory that publish a weekly commitment, 118 — 47% — ask for less than 40 hours a week. The commonest part-time figure is 20 hours, on 84 listings.
Why does the hourly rate make these roles look better paid than they are?
Because an hourly rate is almost always annualised at 40 hours a week, and for these listings that assumption is wrong. Across the 112 part-time roles here that publish both hours and a rate, quoting them at full time overstates the year by a median of $56k — and by $299k at the extreme.
Do shorter-hours roles pay a higher rate?
It looks that way in the data and the causal reading is wrong. The 15-hour band carries the highest median rate here, but those listings are pharma and clinical consulting roles; the 20-hour band is dominated by music, language and creative review work. Specialists are hired for short weeks — short weeks don't pay specialists' rates.
How many of these roles publish their hours at all?
251 of 764. Only one platform in this directory publishes a weekly commitment, so for the remaining 513 roles no figure exists to report. Those are not counted as full time anywhere in this study — every percentage here is taken over the roles that publish, not over the catalogue.
Does part-time mean guaranteed hours?
No. The published figure is what the listing asks for, not a commitment either way. This is contract work paid for time actually worked, so treat the number as the shape of the engagement rather than as hours you can count on.