AI Training Jobs: What Six Companies Say They Pay

In brief

AI training jobs are remote contract work where you write, rank and check AI answers. Six companies state pay on their own pages from up to $17 an hour to a $350 an hour specialist listing, as of November 2026. Most want a degree or real expertise, and the FTC warns against any job that charges you to start.

What are AI training jobs, and what do they pay?

An AI training job is paid work that teaches a model: you write prompts and ideal answers, rank two replies, check facts, or review code and math. The table shows what each company says on its own page, read on 22 November 2026.

A brass balance scale on a black plinth holding two small glass cubes, one clear and one frosted, beside an open notebook with a ribbon marker, next to a terracotta Claude starburst, a brushed silver OpenAI knot and a rainbow glass Gemini star, in a charcoal studio under white rim light.
CompanyPay as stated on its own pageWho it says it wantsWork it describes
DataAnnotationStarting at $25 to $50 an hour (general), $40 to $150+ (coding)General: bachelor's degree or equivalent experience; fluent EnglishTracks: general, multilingual, coding, STEM, professional
OutlierNo hourly rate stated; competitive pay, paid weeklyUndergraduate level expertise typically; no AI experience neededDomain expert tasks
MercorAverage contracted rate $94 an hour; latest listed roles $70 to $350No formal list; experts such as physicians, lawyers, engineersExpert contracts, benchmarks
Alignerr$80 an hour average; roles $40 to $120PhDs, researchers, experts; some roles need no degreeEvaluation, task writing, review
Handshake AIUp to $17 to $150 an hour, by roleMust be based in the U.S.; degree levels vary by roleFellowship projects, part time
OpenTrainTypical $45 to $111 an hour; listings from $7 to $200Most roles entry level; specialist roles need a degree or licenseEvaluation, annotation, coding, translation

Pay on these pages is an average, a starting rate or an up to ceiling, so it is not a promise. Outlier is the only one of the six that states no hourly figure at all. ZipRecruiter and Indeed also list roles with their own pay bands, and the figures vary by listing and city.

Are AI jobs the same as AI training jobs?

No. AI jobs is the wide term, and AI training jobs are one slice of it. The wide term covers engineers who build models, analysts who use them, and trainers who grade what they say. Most AI training work is contract and part time, while engineering roles are usually salaried.

If you searched for ai jobs hoping for a salaried career, AI engineer salary covers that side. If you're worried about the other direction, AI job loss statistics separates measured job cuts from forecasts. This page covers the work that sits between: paid tasks that help a model improve.

What does the work actually look like?

The work is mostly judgement. You read a model's answer and decide whether it is right, you write a harder question than the model can handle, or you fix a code snippet. OpenTrain groups its roles into model evaluation and RLHF, coding and software review, data annotation and collection, transcription and audio, translation and language work, and specialist fields such as law, finance, medicine and engineering.

RLHF, which OpenTrain lists first, stands for reinforcement learning from human feedback: people rate a model's replies, and the model is tuned toward the replies people prefer. That is why careful, honest ratings matter more than fast ones. A rushed rating teaches the model the wrong lesson, and Outlier's page lists quality based rewards next to its pay.

Three real cases from the companies' own pages show how wide the spread is. Handshake AI lists an AI Evaluation Specialist at up to $17 an hour and a Software Engineer at up to $150, about nine times more for the same company. DataAnnotation says its coding track starts at $40 to $150+ while its general track starts at $25 to $50. OpenTrain lists a Question Answer Evaluation Contributor at $7 to $20 an hour and specialist evaluators at $75 to $200, including $150 for financial modeling and $200 for mechanical engineering.

The pattern is the same everywhere. General rating pays the least, and a skill a model can't fake pays the most. The jobs last as long as the projects do. A high rate this month says nothing about steady work next month, and no company on this list promises it. Treat the first project as a trial for both sides.

Who can get an AI training job with no experience or degree?

Almost anyone can apply, but fewer get in. Outlier says "No AI experience needed" and that undergraduate level expertise is typically the minimum. DataAnnotation's general track asks for a bachelor's degree or equivalent experience, strong writing and critical thinking, and fluent English. Handshake AI says no AI experience is required but only takes applicants based in the U.S. who have work authorization it can support.

OpenTrain says most of its roles are entry level and rely on careful reading, clear writing and following guidelines, while specialist roles need a degree, license or professional experience. So the honest answer for ai training jobs with no degree is that some entry roles exist, they pay at the low end, and the high rates sit behind credentials or a skills test.

You'll also face a screen. DataAnnotation says most people spend about an hour on its Starter Assessment and can take it only once. Prepare for it as you would a work sample. Read the task twice, write in full sentences, and explain the reason behind each choice, because the reviewer reads your reasoning as well as your answer.

How can you tell a real AI training job from a scam?

A real employer never asks you to pay to get hired. The FTC's job scam guidance says "Don't pay for the promise of a job," and it lists warning signs such as big pay for little work, fees to be placed, and a check you're told to forward. Its Job Scams page is dated March 2023.

DataAnnotation's own FAQ draws the same line. It says legitimate platforms never charge signup fees or sell training materials, and it names guaranteed income and starter kits as red flags. It adds: "We will never ask for money from you for anything."

Check these six things before you share any personal detail:

  1. The application page is on the company's own domain, which you typed or reached from its site.
  2. No one asked you for a fee, a deposit or a purchase.
  3. The pay is stated as a range or an average, not as a guaranteed income.
  4. No one contacted you first by text or chat app with a job you never applied for.
  5. You were never sent a check to deposit and pass on.
  6. A search for the company's name with "scam" or "complaint" turns up no pattern of the same story.

If you've already paid, the FTC says to contact the company you used to send the money straight away and to report it at ReportFraud.ftc.gov.

What are the downsides of AI training work?

The main downsides are that the hours aren't guaranteed, the pay varies, and the work can end suddenly. DataAnnotation advertises "no minimum hours," and Outlier says "Your earnings are based on the hours you contribute each week." Both lines have a flip side. If projects dry up, your income stops. Pay by the hour also means pay only for the hours a project has.

Three other limits matter. First, "up to" rates describe the best role, not the average one: Handshake AI's top listing is nine times its lowest. Second, you may need to pass a one time assessment before you see any work. Third, platforms differ on who they hire, and Handshake AI limits its fellowship to people based in the U.S.

Before you accept work, ask three plain questions. Are you a contractor or an employee? Who handles your tax? What happens to your access if a project ends? A platform that answers clearly is a better sign than one that dodges.

What can you ask an AI about AI training jobs?

An assistant can help you prepare, but it can't tell you whether a posting is real. These five prompts work in Claude, ChatGPT, Gemini or Whizi, and each asks the model to sort what it knows from what it guesses.

  1. Match my skills: "I have [your degree or work history]. List which AI training tracks, such as general writing, coding, STEM, law, finance or medicine, fit me, and what each would likely ask me to do. Say which parts you're unsure about."
  2. Practice for an assessment: "Give me five short writing tasks like the ones an AI training assessment might use, such as ranking two answers and explaining why. Wait for my reply before you show a model answer."
  3. Check a posting: "Here is a job posting: [paste]. Using only the text, list anything that matches the FTC's job scam warning signs, and anything that is missing, such as a company name or a pay range."
  4. Rate my reasoning: "Here are two AI answers to the question [question] and my ranking with reasons: [paste]. Tell me where my reasoning is weak, and don't just agree with me."
  5. Plan the money: "I'd work [hours] a week at [rate] an hour as a contractor. List the costs a contractor has to plan for, such as taxes and gaps between projects, and say which ones I should check with an accountant."

Treat any answer about a company's pay as a lead, not a fact. Open the company's page, and use how to fact check AI answers for a two minute routine.

Where does Whizi fit?

Whizi doesn't hire anyone, but it helps with the prep: one app instead of a pile of single purpose tools. Pro carries Claude, GPT and Gemini in one app, and its first week is $1.99. Side by side Compare, a Pro feature, puts two models on one prompt, which suits prompt 4 because you can see where they disagree with you and with each other. The free plan gives you 10 messages a day on Whizi AI, Auto and Claude Haiku, with no card.

Web search is on by default in each new chat, so a question about a company's current pay can come back with sources to open. A dedicated job board is better for finding listings, and the company's own page is better for what it pays. For a resume to go with an application, see AI for resumes and cover letters.

Workflow checklist
  • Read the pay on the company's own page and the listing, not a roundup
  • Apply only through the company's own domain
  • Pay nothing, forward nothing, and share no bank login
  • Treat an up to rate as a ceiling, not an average
  • Prepare for the one time assessment before you open it
  • Keep a second source of income while you start
Common questions

Frequently asked questions

Are AI training jobs legit?

Yes, the companies in this guide run real programs, but fake postings copy them. Real employers never charge to hire you. Check the application address, pay nothing and read the FTC's job scam signs before you share personal details.

How much do AI trainers make?

The six companies in this guide state anything from up to $17 an hour to a $350 an hour specialist listing, as of November 2026. Mercor lists an average contracted rate of $94 and Alignerr lists $80 average pay. General tasks pay the least, and coding, law, finance and medicine pay the most.

Do you need a degree for AI training jobs?

Not always, but most platforms screen on one. Outlier typically asks for undergraduate level expertise, DataAnnotation's general track asks for a bachelor's degree or equivalent experience, and OpenTrain says most of its roles are entry level. Specialist roles need a degree, license or experience.

Can AI training jobs be done from home?

Yes. Outlier, DataAnnotation, Alignerr, Handshake AI and OpenTrain all describe the work as remote. Some are limited by country, so check where each one hires before you apply.