Pay by platform (checked 20 Oct 2026)
Every figure below is the platform's own published number, read on its site on 20 Oct 2026 Verified. "Average" and "up to" figures are the platforms' marketing, not offers. We have no affiliate or referral relationship with any of these platforms.
| Platform | Run by | What it publishes | Typical entry requirement | How it pays |
|---|---|---|---|---|
| DataAnnotation | Surge Labs | General $25–$50/hr; multilingual $25–$40; coding $40–$150+; STEM and professional $40–$125+ (FAQ) | Bachelor's or equivalent experience | PayPal, a few days after you request it |
| Outlier | Scale AI | No rates published. "Rates vary by expertise, project complexity, and location. You'll always see the tasking rate before starting any project." (FAQ) | At least an associate degree | Weekly on Tuesdays; PayPal, Airtm or ACH |
| Alignerr | Labelbox | "$80/hr average pay"; listed jobs at $30–$40/hr (English writing), $30–$35/hr (Korean), $40–$120/hr (chatbot testing, writing, AI ethics) (homepage) | Varies by job | Stripe Connect or PayPal |
| Mercor | Mercor | "$107/hr average contracted rate"; oncologist $350–$450/hr, clinicians $120–$150/hr, physicists (PhD) $80–$110/hr, Spanish tutors $60–$65/hr (experts page) | Usually undergraduate-level expertise; many roles need years of experience or an advanced degree | Weekly |
Two notes on those numbers:
- They move. Mercor's average showed $111/hour earlier in October and $107 on 20 Oct. DataAnnotation's multilingual rate read "$20+" on 17 Oct and "$25–$40" on 20 Oct. Rely on the rate shown on each project.
- Location changes pay. Outlier says so directly. Expect lower rates outside the US for the same kind of task.
Lower-paid AI data gigs
Not all AI data work is expert rating. For comparison:
- Prolific studies: researchers must pay at least $8/hour and Prolific recommends $12/hour or more (Prolific pricing, checked 16 Oct 2026) Verified. See Prolific research studies.
- Filming chores to train robots: $3–$15/hour reported by MIT Technology Review and ABC News, with $15/hour tasks seen on DoorDash Tasks by WIRED Self-reported. See Record chore videos to train household robots.
What it adds up to
Hourly rates only matter if there's work. None of the four platforms guarantees hours: DataAnnotation says availability depends on "the demand for your skills and your past performance." Mercor says projects "can be extended, shortened, or concluded early depending on needs and performance." Outlier says project lengths vary "depending on the needs of our customers."
Worked examples Estimate, using 4.33 weeks a month:
| Profile | Rate | Hours/week | Monthly, if work is steady |
|---|---|---|---|
| Generalist, part time | $25 | 6 | about $650 |
| Strong generalist or coder | $40 | 10 | about $1,730 |
| Software engineer on coding projects | $60 | 15 | about $3,900 |
| Licensed clinician (Mercor-style role) | $120 | 8 | about $4,160 |
These are our calculations, not platform data. Real months are lumpier: a project ends, a dashboard goes empty, and the month comes in at a fraction of the table. People who earn steadily usually qualify on two or three platforms at once.
Our research note on AI data annotation and model evaluation has month-by-month ranges, and AI training contracts for licensed professionals covers the expert end.
What decides your rate
- Your track. Coding, STEM, law, finance and medicine pay more than general writing. DataAnnotation's coding track starts at $40/hour against $25 for general work, and tops out at $150+ against $50.
- Your assessment. Every platform screens you. DataAnnotation gives you a single attempt.
- Quality scores. Better scores unlock more and better-paid projects, and poor scores can cut you off.
- Where you live. Outlier says rates vary by location. Alignerr's FAQ lists US ID documents for verification, so check eligibility first.
The fine print
- You're a contractor. All four treat workers as independent contractors: no benefits, and you handle your own taxes. Outlier and Alignerr say they don't sponsor work visas.
- Classification is being challenged in court. Scale AI (Outlier) agreed to a $12.5 million California settlement over contractor classification, with a final approval hearing on 30 Oct 2026 (settlement site). A proposed class action against Surge Labs (DataAnnotation), filed 20 May 2025, is pending, and White v. Mercor.io Corp. (N.D. Tex., filed 8 May 2026) alleges misclassification too (AI Trainer court docket). These are allegations unless a court rules.
- Never pay to work. Fake "AI rating" jobs that ask for a deposit are a known scam (FTC). Real platforms pay you, not the other way round.
Choosing between them? See DataAnnotation vs Outlier vs Alignerr and Is DataAnnotation legit?
FAQ
How much does Outlier pay per hour?
Outlier doesn't publish rates. Its FAQ says rates "vary by expertise, project complexity, and location" and that you always see the rate before starting a project. It pays weekly on Tuesdays.
What is the highest-paying AI training job?
Specialist medical and legal roles. On 20 Oct 2026, Mercor listed oncologists at $350–$450/hour and clinicians at $120–$150/hour. DataAnnotation and Alignerr list some expert roles at up to $150/hour or more.
Can you make a living from AI training jobs?
Some people do, but none of these platforms guarantees hours, and projects end without notice. At $40/hour and 10 hours a week, steady work comes to about $1,730 a month. Treat it as variable side income.
Do AI training jobs need a degree?
Usually. Outlier requires at least an associate degree. DataAnnotation asks for a bachelor's or equivalent experience for general work. Mercor typically wants undergraduate-level expertise, and many of its roles need an advanced degree or years of professional experience.
How do AI training platforms pay?
DataAnnotation pays through PayPal. Outlier pays weekly via PayPal, Airtm or ACH. Alignerr uses Stripe Connect or PayPal, and Mercor says it pays weekly.