AI Training Jobs Pay Report: October 2026
Last updated: 26 October 2026 · OpenJob editorial team
By Antonio Pannella, founder of OpenJob
Data snapshot: 22 October 2026 · 1,047 open listings · Published monthly
Every day we collect remote AI training jobs from the platforms that hire for them and record the pay exactly as each listing states it. On 22 October 2026 there were 1,047 open listings in our database. This report is what those listings say about pay. We did not run a survey or estimate anything: every figure comes from a rate a platform published.
1. Most paid listings start between $30 and $100 an hour
Of the 763 listings that publish an hourly rate, more than half start between $30 and $100. The low end is real too: 136 listings start below $15.
| Starting rate | Listings | Share |
|---|---|---|
| Under $15 | 136 | 18% |
| $15–$29 | 95 | 12% |
| $30–$59 | 193 | 25% |
| $60–$99 | 222 | 29% |
| $100 and over | 117 | 15% |
2. Your field is the biggest pay factor in the data
Professional expertise is what separates high and low rates. The median finance listing starts at $90 an hour and the median healthcare listing at $80. Transcription and data collection start at $10–13.
| Field | Listings with hourly pay | Median start | Median top | Median start |
|---|---|---|---|---|
| Finance | 26 | $90 | $130 | |
| Healthcare & Medicine | 39 | $80 | $125 | |
| Subject Matter Expertise | 344 | $60 | $90 | |
| Coding & Engineering | 79 | $60 | $100 | |
| Coding & STEM | 69 | $50 | $90 | |
| Audio & Speech | 26 | $46.50 | $55 | |
| Search Evaluation | 11 | $16.50 | $16.50 | |
| QA & AI Evaluation | 34 | $15 | $26 | |
| Data Collection | 23 | $13 | $15 | |
| Transcription | 23 | $10 | $24 | |
| Language & AI | 64 | $8 | $65 |
Fields with fewer than 10 listings with hourly pay are not shown.
3. Watch the width of the range
The Language & AI line above looks dismal: a median start of $8 an hour. It is misleading on its own. 41 of its 64 paid listings come from one platform, Meridial (Invisible Technologies), which publishes very wide ranges such as “$6–$65/hour”. Leave Meridial out and the median start for language roles is $20.
This is not a one-off. In 135 listings, the top of the range is at least three times the bottom. A range that wide tells you little about what you will actually be offered. When you see one, ask how the rate is set before you spend time on assessments.
4. Platforms differ in pay and in transparency
| Platform | Open listings | With hourly pay | Median start | Median top |
|---|---|---|---|---|
| Mercor | 293 | 280 (96%) | $65 | $90 |
| micro1 | 301 | 292 (97%) | $50 | $100 |
| Welo Data | 52 | 48 (92%) | $25 | $25 |
| Alignerr (Labelbox) | 11 | 11 (100%) | $25 | $40 |
| Meridial (Invisible Technologies) | 99 | 99 (100%) | $6 | $65 |
| Turing | 226 | 17 (8%) | too few to compare | |
| LILT | 22 | 0 (0%) | no pay published | |
Mercor and micro1 publish an hourly rate on almost every listing, and a large share of their listings are expert roles, which goes a long way to explaining their higher medians. Turing is the opposite case: it had the third-largest number of open listings, but only 8% stated pay.
5. The top of the market
The highest starting rates this month, all as published:
- Oncologist, clinical trials and treatment reasoning (Mercor): $350–$450/hour
- German generalist voice actor (Alignerr): $250–$280/hour
- Applied scientist, management consultant, investment and finance expert, senior software engineer and data scientist roles (micro1): $245–$280/hour each
- Medical doctor, physician expert (micro1): $230/hour
- Cybersecurity research expert, offensive security (Mercor): $200–$250/hour
These are exceptions. They ask for senior, often licensed experience, and a handful of listings do not make a market.
6. Where you live still decides a lot
Number of open listings that explicitly name each country as eligible. Many listings state no country limit, so treat these as a floor rather than a total.
| Country | Listings | Country | Listings |
|---|---|---|---|
| United States | 281 | Nigeria | 20 |
| Mexico | 37 | Kenya | 20 |
| Canada | 35 | United Kingdom | 19 |
| India | 32 | Pakistan | 19 |
| Brazil | 26 | Spain | 19 |
| Italy | 26 | South Africa | 8 |
| Germany | 24 | Philippines | 7 |
| France | 22 | Japan | 22 |
The United States is named in more than seven times as many listings as the next country, Mexico, including many licensed roles in medicine and law that require US credentials.
What we would do as a job seeker
- Start from your profession, not from “AI jobs”. If you have a degree, licence or several years of professional experience, expert roles in your field pay several times more than generalist tasks.
- Read the range, not the headline number. “$6–$65” and “$60–$70” are very different offers.
- Ask about pay before long assessments when a listing publishes none.
- Treat published rates as rates, not income. What you earn depends on how much work is available, and work is rarely guaranteed. Our guide Are AI training jobs legit? covers what to check.
Methodology and limits
- Data: all 1,047 open listings on OpenJob on 22 October 2026, collected from official platform pages. Closed roles are removed daily.
- Starting rate is the lowest figure of a published hourly range and top is the highest; a single rate counts as both. Medians are used rather than averages so a few very high rates do not skew the picture.
- Excluded from pay figures: listings with no pay (263), annual salaries, per-task or one-time payments, and “up to” rates.
- Fields and platforms are as tagged on OpenJob; each listing has one field. Mercor listings appear on OpenJob as “Expert talent network”.
- Limits: these are published rates, not reported earnings, and the snapshot covers one day. Our database includes the platforms we track, not the entire market.
We will repeat this snapshot every month and compare the figures. You may quote them with a link to this page.
Browse the roles behind the numbers: all remote AI training jobs, micro1 jobs, clinical documentation jobs.
Some application links on OpenJob are referral links. This does not affect which listings are included or the figures in this report.