How much does AI training work pay? Rates by field in 2026
AI trainer pay rates by field: what software engineers, doctors, lawyers and other experts earn per hour, what moves rates up, and how payouts work.
Pay is the first question everyone asks about AI training work, and the honest answer is that it varies more than almost any other kind of remote work: from around $15 an hour for general tasks to $150 or more an hour for specialists, sometimes on the same platform. What you earn depends mostly on your field, how deep your expertise runs, and which projects you land. This guide breaks down the typical rates by field, explains how the pay is actually set, and shows how to position yourself at the top of the range.
How AI training pay works
Most platforms pay by the hour for tracked work time, and the rate is set per project, not per person. When a lab needs cardiology questions answered, it funds that project at a cardiology rate; when it needs product descriptions ranked, that project gets a generalist rate. Two experts on the same platform can earn very different amounts simply because they qualified for different projects.
A smaller share of work is paid per task or per accepted answer, which rewards speed, and a few platforms hire for longer contracted or salaried roles. But hourly project work is the norm, and it is the number you should compare when platforms quote you a rate.
Rates are almost always quoted in US dollars, even for experts outside the US, and you work as an independent contractor. The platform pays you directly; the lab never appears in the transaction. Because labs fund projects rather than roles, rates can also shift over time: brand-new projects and newly launched platforms often pay a premium to attract experts quickly, while long-running projects settle toward a stable rate.
Typical rates by field
The ranges below reflect what platforms have been posting for expert AI training work. Treat them as midpoints of a wide distribution: your actual offers will depend on the project, the platform, and how you do on the assessment.
- Software engineers: $60–$150+ an hour. Coding is the deepest and best-funded category. Engineers who can write tests, review pull requests and build evaluation tasks sit at the top. If this is your field, see AI training jobs for software engineers.
- Doctors and nurses: $70–$150+ an hour. Clinical judgment is scarce and heavily regulated, so medical experts command some of the highest rates, especially physicians and specialists. AI training jobs for doctors and nurses.
- Lawyers: $60–$130 an hour. Contract analysis, legal reasoning and jurisdiction-specific questions pay well, with licensed attorneys earning more than general legal writers. AI training jobs for lawyers.
- Finance professionals: $50–$120 an hour. Accounting, valuation and markets expertise is in steady demand as labs train models on financial reasoning. AI training jobs for finance professionals.
- Scientists: $50–$120 an hour. PhD-level researchers in physics, chemistry and biology are hired to create and verify hard science questions. AI training jobs for scientists.
- Math experts: $40–$100 an hour. Competition-level problem writers and graders earn solid rates, though below the clinical and legal bands. AI training jobs for math experts.
- Writers and multilingual experts: $15–$50 an hour. General writing, editing and non-English language work is the most competitive category, so rates start lower — but experienced writers on the right projects do much better.
Notice the pattern: the closer your expertise is to something a lab cannot get cheaply — a medical license, a law degree, production engineering experience — the higher the floor. General skills have a lower floor but far more openings.
What moves you up or down the range
Within your field, several things decide where in the range you land:
- Depth of credential. A practicing specialist outranks a recent graduate; a staff engineer outranks a bootcamp graduate. Platforms verify this during vetting, so it pays to document it.
- Assessment scores. Most platforms gate projects behind a skills test. Scoring well doesn't just unlock the project — on some platforms it puts you in a higher-paying tier for similar work.
- Project type. Creating new evaluation tasks and adversarial testing usually pay more than ranking answers, because fewer people can do them well.
- Quality ratings. Your early submissions are reviewed closely. High quality scores lead to more invitations and, on several platforms, better-paid projects. Rushing through tasks to log hours is the fastest way to stay at the bottom of the range.
- Language and jurisdiction. Experts who can work in a second language, or in a specific legal or medical jurisdiction, often qualify for projects with far less competition.
What this means per month
Hourly rates only tell half the story, because hours are not guaranteed. A useful way to think about it is in bands. At 5 hours a week — a realistic side commitment alongside a full-time job — a $75-an-hour specialist earns around $1,500 a month, while a $25-an-hour generalist earns around $500. At 20 hours a week, those become roughly $6,000 and $2,000 respectively.
The catch is steadiness. Projects start and end, and there are quiet weeks. Experts who treat this as a primary income generally stay on three or more platforms so that when one project's work dries up, another's is ramping up. People who rely on a single platform report the most volatile earnings — which is itself an argument for applying broadly from the start.
It is also worth separating the first month from the steady state. Your first few weeks pay less per hour of effort: assessments take unpaid time, early tasks are slower while you learn each platform's conventions, and invitations to the best projects arrive after you've proven quality. Most experts report that effective hourly earnings rise noticeably after the first month, then plateau.
AI training pay vs other expert side work
How does this compare with the alternatives? Against general freelance marketplaces, expert AI training work usually pays better per hour than writing, design or admin gigs, because the buyer is a well-funded lab rather than a cost-sensitive small business. Against professional consulting in your own field, it usually pays less per hour — but consulting requires finding clients, scoping work and chasing invoices, while training tasks arrive pre-scoped with the infrastructure handled.
The closest comparison is probably expert networks and paid surveys for professionals, which pay similar or slightly higher hourly equivalents but offer far fewer hours. AI training work sits in a useful middle: professional-grade rates with consumer-grade flexibility. For experts who want meaningful extra income without building a client business, that combination is hard to beat.
How platforms differ on pay
Platforms don't just differ in rates — they differ in how pay is structured. Some, like Mercor, target senior specialists with higher hourly rates and weekly payouts. Others, like Outlier, run tiered systems where your rate rises as you complete more work, with weekly pay. Micro1 hires across a wider range of levels through a single AI interview and pays twice a month. The full breakdown is in our platform comparison, including head-to-head pages like Mercor vs Outlier and micro1 vs Outlier.
The practical takeaway: apply to more than one platform. Projects start and stop constantly, and the best-paying project for your profile this month might be on a platform you haven't tried yet.
Getting paid: payouts, contracts and taxes
Payout schedules vary: weekly is common, twice-monthly appears too, and nearly everyone pays by bank transfer or a payment service in USD. Read the payout terms before you start — minimum thresholds and first-payment delays differ.
You are a contractor, not an employee. No benefits, no withholding: in the US that means setting aside estimated tax, and elsewhere it means declaring the income under your local rules for freelance work. Keep records from day one; platforms issue the relevant tax forms, but the bookkeeping is yours.
One rule that never changes: legitimate platforms never ask you to pay to get work. Application fees, training deposits and paid "certifications" are the signature of a scam.
How to earn more as an AI trainer
- Apply to several platforms, not one. Each has its own projects and rates. Being on three platforms triples the chances that a top-rate project in your field is open.
- Treat the assessment like a job interview. It decides your tier. Prepare, take it seriously, and show your reasoning — explained answers score better than bare ones.
- Specialize within your field. "Lawyer" is a commodity; "Delaware corporate litigator" is not. The narrower and harder to replace your expertise, the higher the rate.
- Protect your quality score. Early work is audited. Slow, careful, well-explained submissions compound into better invitations.
- Watch for one-off studies. Platforms periodically run paid interviews and research studies at flat fees that work out to very high hourly equivalents. They go to vetted experts first.
AI training work won't replace a full-time salary for most people, but for experts it is among the best-paid flexible work available: real professional rates, no commute, and hours you control. See which tasks match your background and what they're paying right now.