No AI training jobs listed right now.
Leave your email and we'll tell you when new AI training and data annotation roles are listed:
Every chatbot you use was shaped by people at home rating its answers. AI training jobs pay you to do that: compare two responses and pick the better one, write prompts that trip the model up, check its math or code, label images and audio, or answer questions in your own field so the model can learn from an expert. The companies hiring are a short list — Outlier (owned by Scale AI), DataAnnotation, Mercor, Alignerr (from Labelbox), Handshake AI and micro1 are the names you'll see most.
Searching "data annotation jobs" mostly turns up Reddit threads and reviews arguing about whether any of it is real. It is real, but the details are what decide whether it works for you: the advertised rate is a ceiling, work arrives in projects that can dry up overnight, and almost all of it is 1099 contract work. This page lists the roles we can verify and puts each platform's own pay, payout and onboarding terms side by side, with the date we checked them.
AI training platforms compared
Every figure below comes from the platform's own site or help center, linked in the row. Rates are what the platforms advertise, not what every worker earns.
| Platform | Work | Advertised pay | How you're paid | Getting in |
|---|---|---|---|---|
| Outlier Scale AI |
Generalist AI evaluation (text, voice, image, video); math, finance and other expert tracks | Generalists up to $15/hr; quant finance experts $40–$150/hr (Outlier) | Weekly, by ACH, PayPal or AirTM | Profile, ID check and a skills check; onboarding takes 30–90 minutes |
| DataAnnotation | Chatbot rating, coding, STEM, multilingual and professional work | General $25–$50/hr; multilingual $25–$40; STEM and professional $40–$125+; coding $40–$150+ (FAQ) | PayPal, a few days after you request a payout | One starter assessment (about an hour), one attempt only |
| Mercor | Expert and generalist model-training projects | Listed roles such as Expert Model Trainer at $70–$120/hr; Mercor cites an average of about $100/hr (Mercor) | Weekly, by Stripe Connect or Wise (payments) | A roughly 20-minute AI interview, background check and work trials; 1099 only (policy) |
| Alignerr Labelbox |
Coding, STEM, writing and audio work | Most roles $40–$120/hr; English writing generalists $30–$40/hr (Alignerr) | By the following Friday, through Stripe or PayPal | Resume, ID check and billing setup; some roles add an AI interview or domain test (process) |
| Handshake AI | Fellowships evaluating AI in your field, plus generalist evaluation | Fellows $30–$125/hr; some evaluation roles up to $17/hr (Handshake AI) | Per project terms | Application, then project-specific steps; onboarding is unpaid unless the project says otherwise |
| micro1 | Expert evaluation and AI agent work, about 15–20 hours a week | No public rate card | Twice a month in USD, through Deel (FAQ) | AI interview, Deel identity check and a background check |
Rates and terms change often. We re-check this table monthly; if a platform's page disagrees with it, the platform's page wins.
Types of AI training work
Generalist rating and writing. You compare model answers, rate them against a long guideline, and write better responses. The entry point for most people, and the lowest-paid track — Outlier advertises up to $15 an hour for generalists, DataAnnotation $25–$50 for general work.
Coding. You review, debug and write code so models learn to program. Expect a coding test; pay is the highest outside the expert tracks, $40–$150+ an hour on DataAnnotation.
Expert and STEM. Professionals and graduate students write and grade hard questions in their field — medicine, law, finance, physics, chemistry. Mercor, Handshake AI's fellowships and Scale AI's Human Frontier Collective fellowships recruit here, generally at $30–$150 an hour depending on the field.
Language. Rating and writing in languages other than English, or checking translations. DataAnnotation lists $25–$40 an hour for multilingual work.
Data labeling. Tagging images, audio or video so a model can learn to recognize them. More repetitive, usually lower-paid, and the closest to classic "data annotation."
How to get accepted
- Pick the track that matches what you can prove. A degree, a professional license, a coding portfolio or a second language moves you out of the generalist pool and into higher-paid work.
- Treat the assessment like an exam. DataAnnotation gives you one attempt at its starter assessment. Read every instruction, write complete answers and take your time — speed doesn't score, accuracy does.
- Prepare for an AI interview. Mercor and micro1 screen with a recorded AI interview, and Alignerr adds one for some roles. Quiet room, working camera, concrete examples from your field.
- Expect identity and background checks. Outlier and Alignerr verify ID; Mercor and micro1 run background checks. That's a sign of a real platform, not a red flag — as long as they never ask you for money.
- Apply to more than one platform. Work comes in projects, and any single dashboard can go quiet for weeks. People who earn steadily usually keep two or three accounts active.
What the advertised rates don't tell you
- Work is not guaranteed. Platforms assign projects based on demand and your past quality scores. DataAnnotation's own FAQ says available work depends on both. Worker reviews on every platform describe empty dashboards after onboarding and sudden removals from projects.
- Some time is unpaid. Assessments, interviews and project onboarding are usually unpaid; Handshake AI says onboarding is unpaid unless the project terms say otherwise.
- It's 1099 income. No one withholds tax for you, there are no benefits, and you'll owe self-employment tax. Our W-2 vs 1099 calculator shows what a contract rate is worth next to a regular job.
- Rates can change mid-project. Workers on several platforms report pay cuts and new time-tracking rules after they started. Screenshot the rate when you accept a project.
For a wider list of platforms beyond the big names, see our guide to 8 remote gigs like DataAnnotation.
How to spot an AI training job scam
No legitimate AI training platform charges you to apply, to train, or to unlock work. If anyone asks for money, it's a scam. DataAnnotation publishes a trust and safety notice warning that people use its name to commit fraud, that it will never ask you for money, and that the only way in is its own sign-up.
- A "recruiter" texts or messages you on WhatsApp, Telegram or Teams offering $25–$40 an hour with no application. Real platforms have you apply on their own website.
- You're asked to pay for training, equipment, a "processing fee" or a background check.
- Someone offers to sell you a "verified" account. AlgorithmWatch reported in 2025 that such accounts are sold in WhatsApp and Facebook groups; platforms deactivate them, and the money is gone.
- A look-alike website or email domain that's one letter off the real one.
If you've been targeted, the FTC's job scams guide explains what to do, and you can report it at ReportFraud.ftc.gov. We only list roles posted on a platform's own site or hiring board.
More ways to work remotely
Guides and explainers
AI training jobs — FAQ
What are AI training jobs?
AI training jobs are remote contract work that helps AI models learn. You rate and compare chatbot answers, write and check prompts, label data, or solve coding and subject-matter problems that the model is then trained on. Outlier (Scale AI), DataAnnotation, Mercor, Alignerr and Handshake AI are among the companies that hire for it.
How much do AI training jobs pay?
Advertised rates as of October 2026: generalist work runs from up to $15 an hour at Outlier to $25–$50 at DataAnnotation, coding work $40–$150+ an hour, and expert or PhD-level work roughly $40–$150 depending on the field. These are advertised ranges, and your real earnings depend on how many projects are available to you.
Are AI training jobs legit?
The major platforms are real companies that pay contractors, but the work is project-based and can stop without notice, and assessments and onboarding are often unpaid. Scammers also impersonate these platforms. Apply only through the platform's own website and never pay a fee to get work.
Do I need a degree for AI training jobs?
It depends on the track. Generalist and language work mostly test your writing and how carefully you follow guidelines, though DataAnnotation asks for a bachelor's degree or equivalent for general work. Expert tracks look for a master's, a PhD or professional experience in the field, and coding tracks look for real programming experience.
Are AI training jobs W-2 or 1099?
Almost always 1099 independent contractor work. Mercor states it does not offer W-2 positions, and Handshake AI fellows are 1099 contractors. Nobody withholds tax for you, so set money aside for income and self-employment tax.
Is AI training work worth it?
As flexible side income, often yes: the work is fully remote, you set your hours, and skilled tracks pay well above most remote entry-level jobs. As a sole income, it's risky — projects end without warning, the rate on the homepage is a ceiling, and there's no employer covering taxes or benefits. Apply to two or three platforms in the track that fits your skills, keep records of every rate you accept, and never pay anyone to get started.