How Much Does Training AI Actually Pay in 2026? Real Listed Rates
I actually work on these platforms — opinions come from my own dashboards and payouts. Some links are referral links that pay me at no cost to you. Full disclosure.
Anywhere from $6/hr to $300/hr with an average around $60/hr for domain experts.
Pay by tier
| Tier | Typical range | Who gets it |
|---|---|---|
| Surveys, transcription & data collection | $6–$40/hr | No credentials needed |
| Generalist evaluation | $20–$65/hr | Good writing, careful reasoning, reliability |
| Expert review (incl. STEM PhDs) | $65–$90/hr | Verifiable professional experience |
| Specialist (medical/legal/finance) | $100–$180/hr | Licenses or deep domain depth |
| Premium specialist | $150–$300/hr | Scarce, region-specific, or niche-market expertise |
The top tier exists — I’ve seen a $300/hr precious-metals-markets role listed — but most hiring happens in the middle. Roles hiring hundreds of people a month cluster between $60 and $125; the $200+ listings are niche and hire in small numbers.
Pay by platform
- micro1: $6–$20/hr transcription and data entry, $20–$45/hr annotation and evaluation, $40–$95/hr science PhDs and rare-language experts, $80–$230/hr top legal, finance, and senior engineering roles. You propose your own rate.
- Mercor: $10–$40/hr survey roles, $65–$90/hr most PhD/STEM expert work, $100–$180/hr legal, finance, and medical, $150–$300/hr premium specialist roles. Weekly payouts.
- Handshake AI: $75–$125/hr Fellowship roles, typically $90–$100/hr, with defined weekly hours.
Full head-to-head: Outlier vs micro1 vs Mercor vs Handshake AI.
What moves you up a tier
- Verifiable credentials — a license, degree, or title a platform can check beats any self-description.
- A work history an assessment can probe — Mercor’s per-skill assessments dig into real applied experience.
- Specialist scarcity — multilingual ability, niche engineering domains, clinical experience, and region-specific professional credentials are chronically undersupplied. On Mercor, a UK employment lawyer lists at 3x the rate of a STEM PhD; on micro1, a Tongan bilingual expert ($45–$95/hr) lists at 4x Spanish transcription ($10–$20/hr).
What the “make $4,000/week” posts leave out
- Volume fluctuates. Every platform has dead weeks; none guarantee hours.
- You’re a contractor. No benefits, and you owe self-employment tax on earnings.
- The window may not last. Demand for human training data is high in 2026, but the sector is young and consolidating. Earn while it’s strong; don’t build your life on it.
Get started: Handshake AI · Mercor · micro1 · Outlier (pick a project)
Frequently asked questions
How much can a beginner make training AI?
Entry annotation and generalist evaluation work typically pays $16–$40/hr. With strong writing and judgment, evaluator roles reach $65/hr without specialist credentials.
Can you make $100/hr training AI?
Yes, but it requires demonstrated expertise: medical, legal, finance, or senior technical backgrounds routinely command $100–$180/hr on Mercor and micro1's expert tiers, and scarce niche specialties list as high as $300/hr.
Is AI training work full-time income?
Usually not. Task availability fluctuates on every platform and none guarantee volume. Most people should treat it as supplemental income of variable weekly hours.