LIVE 2,000+ jobs added today  ·  $20–$200/hr  ·  Apply direct Browse now →
Launch preview · Illustrative until production index ships
// The HireFeed Pay Index  ·  Updated weekly

What AI training actually pays.

The only public rate index built from pay-disclosed job listings — not surveys, not guesses. Every domain shows a typical hourly range, a top-decile rate, the platforms hiring most actively, and a 90-day trend. Neutral. Methodology below.

This is the page every "how much does Outlier pay" or "AI trainer salary" search should land on.

7 Domains tracked
Weekly Refresh cadence
Trailing 30d Sample window
Pay-disclosed Source only
Domain Typical range Top decile Most active platforms 90-day trend
Medicine · MD/DO clinical reasoning · diagnostic eval $100–$210/hr $250/hr Surge · Mercor · Handshake ▲ +12%
Law · JD contract reasoning · compliance $90–$160/hr $250/hr Mercor · Micro1 · Surge ▲ +9%
Math & Science · PhD reasoning · proof verification $60–$120/hr $150/hr Outlier · Handshake · xAI ▲ +15%
Agentic Eval & Red-team trajectory review · adversarial $70–$130/hr $160/hr Labs direct · Mercor ▲ +21%
Coding Evaluation code review · SxS · debugging $40–$75/hr $95/hr Outlier · Alignerr · Micro1 ▲ +7%
Multilingual & Linguistics 40+ languages · localization QA $25–$45/hr $60/hr Surge · Toloka · Outlier ▲ +3%
Generalist Annotation preference ranking · basic QA $18–$25/hr $32/hr DataAnnotation · Outlier ▼ −4%
// How this is computed

The methodology, on one page.

The reason the current SERP for "how much does Outlier pay" is anecdote soup: nobody publishes their method. We do. Everything below is what a reader (or an AI Overview) can cite by name.

// Source
Pay-disclosed listings only. Job postings that carry an explicit hourly rate, salary, or rate band. Postings without disclosed pay are excluded — they can't be verified against a rate.
// Sample window
Trailing 30 days. Long enough to smooth weekly campaign fluctuations; short enough that a genuine market move shows up before the quarter ends.
// Typical range
The middle 50% of disclosed rates for the domain in the trailing window — the range you should actually expect. Not the ceiling, not the floor.
// Top decile
The 90th percentile hourly rate — what the best-paid listings in the category look like. Useful ceiling for salary negotiation.
// 90-day trend
Percent change of the median rate vs. the prior 90-day window. Rising = supply is tightening, or the domain is on a lab's roadmap; falling = task is being commoditized.
// What we don't do
No survey data (self-report inflation is huge). No affiliate weighting — platforms don't pay for placement here. No off-platform anecdotes. If a rate isn't in a job listing, it isn't in the index.
// FAQ

Common questions.

The answers we saw missing from Page 1 for "AI trainer pay rates" and "AI training salary."

How much do AI trainers actually get paid in 2026?

AI training work spans a wide pay range depending on domain and credential requirements. Generalist annotation pays $18–$25/hr. Coding evaluation pays $40–$75/hr. Math and science reasoning (PhD-level) pays $60–$120/hr. Agentic evaluation and red-teaming pays $70–$130/hr. Law (JD) pays $90–$160/hr. Medicine (MD/DO) pays $100–$210/hr. Top decile rates reach $250/hr in medicine and law. All figures are from pay-disclosed listings, not surveys.

Which AI training platforms pay the most?

By domain: Surge, Mercor, and Handshake are most active in medicine and law. Outlier, Handshake, and xAI show the most PhD-level math and reasoning postings. Mercor and labs-direct are where the highest-rate agentic evaluation and red-teaming shows up. Outlier, Alignerr, and Micro1 are the most active in coding evaluation. Surge, Toloka, and Outlier for multilingual. DataAnnotation and Outlier for generalist annotation. Pay bands vary within each platform by task type and worker tier.

Are these rates verified? Where do they come from?

Every figure is drawn from a pay-disclosed job listing that we indexed in the trailing 30 days. They are not survey responses. During launch preview the numbers are illustrative until the production index ships with live listing counts per row; from that point every figure will be dated and traceable to a listing count and a source URL.

Why do agentic evaluation and red-teaming rates rise the fastest?

Frontier labs pushed hard into tool-using agents in the first half of 2026, and the eval methodology for multi-step agent trajectories is scarcer than single-turn preference ranking. Demand rose faster than the pool of experts who have both a domain and rubric-calibration skill, so the 90-day trend is stronger for this category than for adjacent ones.

Do I need a PhD to earn premium rates?

A credential helps but is not the only path. Medicine, law, and PhD-level math and science reasoning pay $60–$210/hr because the work explicitly requires the credential. Agentic evaluation and coding evaluation pay $40–$130/hr and screen on demonstrated skill more than degree. Passing a rigorous rubric-calibration assessment — measured on inter-rater agreement and golden-set accuracy — is a stronger predictor for those roles than the CV.

How often is the Pay Index updated?

The production Pay Index refreshes weekly from the trailing 30-day window of pay-disclosed listings on HireFeed. The 90-day trend column reports the percent change of the median rate vs. the prior 90 days. Every figure is stamped with an update date on this page.

Do you take affiliate money from these platforms?

No. Platforms cannot pay for a higher rank or a better-looking trend on this page. If that ever changes, we will disclose it on this page and any affected figure will carry a "sponsored" mark. The neutrality of the index is the whole point.

// Ready to find one?

2,000+ rate-disclosed jobs, added today.

Every listing on HireFeed's job feed shows the pay upfront. Apply direct — no forms, no data resold, no résumé middleman.

Browse jobs