Skip to main content
StrideRecruit

Finding jobs and networking

Remote AI Trainer Jobs (2026): Pay & How to Start

Remote AI trainer jobs rate chatbot answers and write prompts so models improve. DataAnnotation's FAQ starts general work at $25–$50 an hour.

Remote AI trainer jobs are paid contract work improving chatbots by writing prompts, ranking answers and grading outputs β€” the method behind OpenAI's InstructGPT paper. DataAnnotation's FAQ starts general work at $25–$50 an hour, coding at $40–$150+; Outlier, Mercor, Remotasks and Handshake AI hire the same way. Hours are never guaranteed, and no real platform asks you to pay.

By StrideRecruit Editorial Team Β· Last reviewed: Β· General information, not legal or immigration advice.

Guides are checked against the official sources listed at the end of each guide.

What the work pays

DataAnnotation publishes starting rates by track in its FAQ, and those are the only platform-wide pay bands this guide quotes: general projects start at $25–$50 an hour, multilingual projects at $25–$40, coding at $40–$150+ and STEM and professional projects at $40–$125+. The top of each band is for the hardest work β€” senior engineers on coding, licensed professionals on law, finance or medicine β€” not the rate every contributor sees.

Read β€œstarting at” as the price of an hour of approved project work, not a salary. DataAnnotation says there are no minimum hours and no daily login requirements, and that the work you are offered depends on demand for your skills and your past performance. A week with no matching project pays nothing, however good the hourly rate looks.

Other platforms show the rate per role, not per platform. On Mercor's homepage on 16 November 2026, individual roles were listed at $150 an hour for a General Physician and $60–$70 an hour for a Sales and Marketing Expert β€” rates for those roles only, not a Mercor-wide band. Handshake AI posts rates upfront: one fellow on its program page says rates β€œare posted upfront. They never move, never change.” Outlier describes its pay only as competitive, with quality-based rewards paid weekly, so check the rate on the project itself before you start.

Starting pay (DataAnnotation FAQ, Nov 2026)
General projects$25–$50 per hour
Multilingual projects$25–$40 per hour
Coding projects$40–$150+ per hour
STEM projects$40–$125+ per hour
Professional projects$40–$125+ per hour

Source: DataAnnotation: FAQs (starting rates by track, Starter Assessment, PayPal); Mercor: Organizing human intelligence to power the AI economy (role rates shown on homepage)

What the work actually is

The method behind this work was described in OpenAI's 2022 InstructGPT paper: human labelers wrote demonstrations of the behaviour a model should show, then ranked the model's outputs, and both datasets were used to fine-tune the model with reinforcement learning from human feedback (RLHF). AI training jobs are still built around those two activities β€” showing the model good work, and ranking its attempts.

In practice the tasks are: writing hard prompts in your own field, with the correct answer worked out; creating grading rubrics that say what a good answer must contain; rating and ranking AI answers; and reviewing AI-generated code, maths or specialist content for errors. DataAnnotation's FAQ describes the same day-to-day work: evaluating chatbot responses for accuracy, comparing AI outputs, flagging factual errors, testing AI-generated images, writing challenging prompts and reviewing code.

At the simpler end sits data annotation β€” labelling raw data so a model can learn from it. Remotasks' tasks are transcribing text and audio, labelling images and labelling LiDAR data, and it says LiDAR and segment annotation tend to pay the most and have the steadiest supply. Annotation needs care and consistency more than a degree; the expert tracks need the subject knowledge.

Source: Ouyang et al.: Training language models to follow instructions with human feedback (InstructGPT, arXiv:2203.02155); DataAnnotation: FAQs (starting rates by track, Starter Assessment, PayPal); Outlier: Train the Next Generation of AI as a Freelancer (tasks, onboarding, qualifications)

Who hires AI trainers

Five platforms hire for this work, all verified on their own public pages on 16 November 2026. DataAnnotation recruits remote contractors to rate, write and correct AI responses, with separate tracks for general, multilingual, coding, STEM and professional work. Outlier, operated by Scale AI, connects experts with leading AI companies to provide the human feedback that improves large language models. Remotasks supplies training data to AI companies, mostly in the self-driving car space, through short paid tasks. Mercor runs a network of experts who evaluate AI and build evaluations against professional standards. Handshake AI hires experts and generalists for paid remote AI training work, as its program page puts it.

The platforms split roughly two ways. DataAnnotation, Outlier and Handshake AI are expert-led: they match your subject β€” maths, code, law, medicine, languages β€” to projects that need it. Remotasks is task-led: you train on a task type, get enabled for it, and take tasks as they appear. Mercor sits between the two, listing individual expert roles with their own rates.

None of these is a conventional employer. DataAnnotation describes its workers as contractors paid through PayPal, Outlier calls its people freelance experts, and Handshake AI calls its workers fellows on project contracts. That means no fixed hours, no guaranteed workload and, in most cases, no benefits β€” the trade-off for choosing when and how much you work.

Source: DataAnnotation: FAQs (starting rates by track, Starter Assessment, PayPal); Outlier: Train the Next Generation of AI as a Freelancer (tasks, onboarding, qualifications); Remotasks: Frequently Asked Questions (sign-up, training courses, tasks, no fee to join); Mercor: Organizing human intelligence to power the AI economy (role rates shown on homepage); Handshake AI Program: Paid, Remote AI Training Work for Experts and Generalists

What you need to qualify

For general work the bar is writing, not credentials. DataAnnotation's baseline is a bachelor's degree or equivalent real-world experience, with strong writing and critical thinking, and no coding experience is required for its general projects. Outlier says no AI experience is needed and asks for undergraduate-level expertise in your domain, preferring graduate students and master's or PhD holders. All tracks require fluent English and reliable internet access.

The higher bands ask for proof. On DataAnnotation, STEM projects need a master's or PhD in maths, physics, biology or chemistry β€” or a bachelor's degree plus 10 or more years of professional experience β€” and professional projects need licensed credentials in law, finance or medicine. Coding projects need programming experience in languages such as Python or JavaScript, including solving LeetCode-style problems.

You will also need to prove who you are. DataAnnotation verifies identity through Persona, its verification partner, with a physical government-issued photo ID and a selfie. Outlier asks for a valid ID and mobile phone from your country of residence, a current resume and a LinkedIn profile showing your education and work experience. Keep those ready before you apply β€” a half-finished profile waits longer.

Source: DataAnnotation: FAQs (starting rates by track, Starter Assessment, PayPal); Outlier: Train the Next Generation of AI as a Freelancer (tasks, onboarding, qualifications)

How to apply, step by step

On DataAnnotation you sign up, choose your qualification track, and take the Starter Assessment β€” about an hour for the general track, one to two hours for specialist tracks in coding, maths, science, law, finance, medicine or a language. You get one attempt: the FAQ says there are no retakes or second chances, so read the instructions carefully and review your answers before submitting. If you pass, you are told by email within a few days and can start choosing projects.

On Outlier, general onboarding typically takes 30 to 90 minutes: create your account, select your areas of expertise, complete skill screenings so you can be matched with the right projects, and verify your identity. After that comes project-specific onboarding, which varies by project. There is no traditional interview on either platform β€” the assessment is the interview.

Remotasks works differently: sign up with Facebook or Gmail, take courses in its Training Center, and once you complete a course and are enabled to work on a project, you can start doing tasks and earning. Take more courses to get enabled on more projects β€” LiDAR and segment annotation first, since Remotasks says those pay the most and have the steadiest supply. Mercor and Handshake AI both take applications on their own sites β€” Mercor at work.mercor.com, Handshake AI at joinhandshake.com/ai β€” where each role or project lists its own requirements and rate.

Source: DataAnnotation: FAQs (starting rates by track, Starter Assessment, PayPal); Outlier: Train the Next Generation of AI as a Freelancer (tasks, onboarding, qualifications); Remotasks: Frequently Asked Questions (sign-up, training courses, tasks, no fee to join); Mercor: Organizing human intelligence to power the AI economy (role rates shown on homepage); Handshake AI Program: Paid, Remote AI Training Work for Experts and Generalists

What the work is really like

Expect project work, not a job with a rota. DataAnnotation and Outlier describe the arrangement the same way: projects run around the clock, you choose which to take, and there are no minimum hours. DataAnnotation says its projects run 24 hours a day, 365 days a year, with no minimum hours and no daily login requirements. Outlier says the same β€” you choose when, where and how much you contribute, with no minimum hour requirements.

The honest catch is availability. DataAnnotation says plainly that the work you see depends on demand for your skills and your past performance, and that poor-quality contributions can cost you access to future projects. Treat this as variable extra income, not a salary: some weeks will have plenty of work, others almost none, and the hourly rate only counts for hours on an approved project.

Quality is the currency. Both DataAnnotation and Outlier screen on quality, reward it with higher rates and better projects, and remove contributors whose work falls short. Read the project guidelines before every task, follow them exactly, and check your own answers the way a reviewer would β€” that is the whole skill, and it is what the assessments are testing.

Source: DataAnnotation: FAQs (starting rates by track, Starter Assessment, PayPal); Outlier: Train the Next Generation of AI as a Freelancer (tasks, onboarding, qualifications)

Where to apply

Start with the live remote jobs linked below, and use the remote filter before you read anything else. The list updates as new roles arrive, and every listing links to the original posting, where you apply.

Check each posting for which countries and time zones it covers before you apply β€” AI training platforms often limit projects by country, and a role open in one country may not be open in yours. Apply through the platform's own site, and keep a copy of the posting text.

If a platform has no live role that fits you, join it anyway and complete its assessment: DataAnnotation and Outlier both match you to projects after you pass, and new projects appear as AI labs order them. A listing on any board is not a guarantee the role is real β€” apply the scam checks in the next section to every posting, on any site.

Scam safety: the rules that never change

A real job never asks you to pay. The US Federal Trade Commission puts it plainly: honest employers, including the federal government, will never ask you to pay to get a job. DataAnnotation says the same in its FAQ β€” β€œWe will never ask for money from you for anything” β€” and Remotasks says it does not charge for working on its platform. Anyone who asks for a fee β€” for training, equipment, a background check, a starter kit or β€œunlocking” your earnings β€” is running a scam.

The common versions in AI work: a β€œrecruiter” who contacts you on WhatsApp or Telegram about an AI trainer role and asks for payment or personal documents; a lookalike site with a name close to a real platform; and promises of high pay for little work. Losses to job scams topped $220 million in just the first half of 2024, according to the Federal Trade Commission.

Apply only by typing the platform's own address yourself β€” dataannotation.tech, outlier.ai, remotasks.com, mercor.com, joinhandshake.com β€” and never through a link in a message. If an offer shows any of these signs, stop, keep screenshots of the advert and the messages, and report it to the Federal Trade Commission at ReportFraud.ftc.gov β€” and to the site where you saw the listing.

Source: US Federal Trade Commission: Job Scams; US Federal Trade Commission: Job scams drove record losses in early 2024; DataAnnotation: FAQs (starting rates by track, Starter Assessment, PayPal); Remotasks: Frequently Asked Questions (sign-up, training courses, tasks, no fee to join)

Sources

Frequently asked questions

How much do remote AI trainer jobs pay?
DataAnnotation's FAQ, read 16 November 2026, starts general projects at $25–$50 an hour, multilingual at $25–$40, coding at $40–$150+ and STEM and professional projects at $40–$125+. Those are starting rates for approved project hours, not a salary β€” there are no minimum hours, and the work you are offered depends on demand for your skills. Mercor shows the rate per role instead: $150 an hour for a General Physician and $60–$70 for a Sales and Marketing Expert on its homepage the same day.
What does an AI trainer actually do?
You teach AI models with human feedback: writing hard prompts with the correct answer, creating grading rubrics, rating and ranking AI answers, and checking AI-generated code, maths or specialist content for errors. The method comes from OpenAI's 2022 InstructGPT paper, where labelers wrote demonstrations of good behaviour and ranked model outputs to fine-tune the model. Simpler annotation work β€” labelling images, transcribing audio, labelling LiDAR data β€” is the entry level.
Do I need a degree to train AI?
For general work, no degree is strictly needed: DataAnnotation asks for a bachelor's degree or equivalent real-world experience, and Outlier asks for undergraduate-level expertise in your domain, with no AI experience required. The higher-paying tracks do need proof β€” a master's or PhD, or a bachelor's plus 10 or more years of experience, for STEM, and licensed credentials for law, finance or medicine.
Which platforms hire AI trainers?
Five, all with public hiring pages: DataAnnotation, Outlier (operated by Scale AI), Remotasks, Mercor and Handshake AI. DataAnnotation, Outlier and Handshake AI match your subject expertise to projects; Remotasks trains you on task types such as image labelling and LiDAR annotation; Mercor lists individual expert roles with their own rates. All five are free to join.
How do I apply to DataAnnotation?
Sign up on dataannotation.tech, choose your qualification track, and take the Starter Assessment β€” about an hour for the general track, one to two hours for specialist tracks. You get one attempt, with no retakes, so review your answers before submitting. If you pass, you are told by email within a few days and can start choosing projects. You will also verify your identity with a government photo ID.
Are remote AI trainer jobs full-time work?
No. It is project-based contractor work with no minimum hours: you choose which projects to take and when to work, and projects end when the AI lab's need ends. DataAnnotation says the work you are offered depends on demand for your skills and your past performance, so treat it as variable extra income, not a salary. Some weeks will have plenty of work and others almost none.
How do I spot an AI trainer job scam?
A real platform never asks you to pay β€” DataAnnotation's FAQ says it will never ask you for money, and the FTC says honest employers will never ask you to pay to get a job. Apply only on the platform's own site, never through a link in a WhatsApp or Telegram message, and beware of lookalike domains and promises of high pay for little work. Report scams to the FTC at ReportFraud.ftc.gov.

Browse live jobs

See current openings on StrideRecruit.

Explore 28,413 Jobs β†’