What AI training pays.

General tasks such as annotation and transcription list a median $27/hr across 229 rates in October 2026. Expert roles list $77.50/hr across 2,273.

All 2,502 rates together give a median of $75/hr. Counting each platform once gives $46/hr, across the 11 platforms with 10 or more rates.

The sample covers 16 platforms on October 6, 2026: Ethos, micro1, SME Careers, Mercor, Turing, AfterQuery, Vetto, Sovrano, xAI, Innodata, RWS, Outlier, Handshake, Mindrift, Meridial, and Alignerr.

Median listed rate · October 6, 2026
Group Median Middle half Rates (n)
General tasks $27/hr$20 to $35 229
Expert roles $77.50/hr$55 to $100 2,273
All $75/hr$48 to $100 2,502
Each platform counted once $46/hr from each platform's own median 11 platforms
Median rate by kind of work · October 6, 2026
Kind of work Median Middle half Rates (n)
Psychology $122/hr$70 to $150 29
Healthcare $100/hr$75 to $150 164
Legal $100/hr$75 to $125 151
Finance $100/hr$80 to $110 91
Accounting $85/hr$80 to $105 37
Business $80/hr$70 to $100 338
STEM $77.50/hr$60 to $95 330
Software Engineering $75/hr$50 to $100 180
Humanities $75/hr$55 to $80 62
Mathematics $75/hr$50.50 to $95 35
Generalist $70/hr$50 to $100 436
Design $70/hr$52.50 to $90 51
Voice Acting $50/hr$41 to $75 43
Writing $45/hr$22.50 to $76 111
Data Science $45/hr$15 to $76 92
Languages $35/hr$22.50 to $50 345
Creative under 10 rates 4
IT & Systems under 10 rates 4
QA Testing under 10 rates 1
Median rate by platform · October 6, 2026
Platform Median Middle half Rates (n)
AfterQuery $87/hr$70 to $115 172
Handshake $80/hr$75 to $100 206
Ethos $80/hr$70 to $80 203
Mercor $75/hr$50 to $100 1,360
micro1 $70/hr$46 to $137 231
Mindrift $46/hr$33 to $55 103
xAI $35/hr$35 to $35 32
Alignerr $22.50/hr$22.50 to $35 60
Meridial $22/hr$20 to $60 47
RWS $15/hr$11 to $18 37
Innodata $15/hr$15 to $37.50 27
Sovrano under 10 rates 9
Outlier under 10 rates 7
Vetto under 10 rates 5
SME Careers under 10 rates 2
Turing under 10 rates 1

Download the tables (CSV) Download the tables (JSON)

Cite as: AI Training Board Pay Report, October 2026, aitrainingboard.com/pay/2026-10. Free to reuse with credit, under CC BY 4.0.

Rules version 2026-10. The version changes whenever a rule below changes, and the CSV and JSON carry it.

How the figures are computed

We take each open role's rate as the platform lists it, on October 6, 2026, from 3,488 open listings on 16 platforms. Only hourly rates count. We leave out our own estimates, rates a listing gives only as a ceiling ("up to $50/hr"), ranges whose top is more than 10 times the bottom, and rates under $3 or over $400/hr. A role posted twice with the same title and rate counts once. A listing that gives a range ("$80 to $150/hr") counts at its midpoint in these medians; the board itself shows the range.

A general task is a role whose title names work anyone can do, such as annotation, labeling, transcription, video or voice recording, rating AI responses or bilingual generalist work, and carries no profession, field or seniority word. Every other role is an expert role. The split reads the title only, so a role titled "Generalist Annotator" is a general task and "Radiologist" is not, whatever the platform calls its category.

One platform supplies over 40% of all rates, so the median of all rates together mostly describes that platform. To count each platform once, we take each platform's own median, then the median of those, over the 11 platforms with 10 or more rates.

A group needs 10 different rates before we show a median. The middle half is the range between the 25th and 75th percentile, so a quarter of the rates sit below it and a quarter above. Listed rates are what a platform advertises, so what you're paid can differ by project and by country. New figures go up on the 1st of each month, and each month keeps its own page.

Earlier months