Data Annotation Jobs: Who Hires, What They Pay, and Is DataAnnotation Legit? (2026)

The short answer: data annotation jobs pay you to label or judge data so AI systems can learn from it. That can mean drawing boxes around objects in photos, transcribing audio, tagging video, rating search or ad results, or evaluating chatbot answers. Most of the work is remote, flexible and paid by the hour as a contractor, through platforms such as DataAnnotation, CrowdGen (Appen), TELUS Digital’s AI Community, Outlier and Mindrift. DataAnnotation itself is a real platform that pays through PayPal, and says general projects start at $25 to $50 an hour. But you get one attempt at its Starter Assessment, and reporting has documented workers who passed and never got tasks, or lost their accounts with earnings still unpaid. So treat any single platform as one income source among several, and never pay anyone to get the work.

Platform details below come from each platform’s own pages and from published reporting, as of 8 November 2026, with sources at the end.

Disclosure: Huru is made by Beatview, Inc., which also makes Beatview, an AI recruiting agent that recruitment agencies use to run AI video interviews. We’re not affiliated with any platform named here and don’t earn anything from these links.

What data annotation work involves

Annotation turns raw data into examples a model can learn from. The common task types are:

• Image and video labeling: drawing bounding boxes or outlines around objects, tagging what’s in a frame, or marking actions in a clip.
• Audio work: transcribing speech, marking who’s speaking, or rating recordings.
• Text classification: tagging sentiment, topic or intent, or flagging unsafe content.
• Search and ad evaluation: judging whether results or ads match what someone searched for, often in your own language and country.
• Chatbot evaluation: rating AI answers for accuracy, helpfulness and safety, and explaining your rating.

The last type overlaps with AI trainer work, which leans more on subject expertise: writing hard prompts, model answers and grading rubrics in fields like medicine, law or coding. If that’s your background, our guide to AI trainer jobs covers the expert platforms and their pay in detail.

Who hires data annotators

What each platform says about the work:

• CrowdGen is Appen’s contributor platform. It describes the work as evaluating, annotating, generating and refining data, says most projects don’t require technical skills, and lets you choose projects and set your own pace. You create an account, pass a verification review and often a project qualification, and CrowdGen says it will never ask you to pay to join a project.
• TELUS Digital’s AI Community hires part-time remote evaluators across many countries and languages, in roles such as Internet Safety Evaluator, Personalized Internet Assessor and Online Data Analyst. That makes it worth a look if you live outside the US or speak a language other than English.
• Outlier and Mindrift focus more on AI training tasks; see our AI trainer jobs guide for both.

One platform you’ll still see recommended in older lists is gone: Amazon Mechanical Turk says it permanently closed on 30 September 2026.

Is DataAnnotation legit?

Yes, it’s a real platform that pays people, but go in with clear expectations. Here’s what DataAnnotation says, and what independent reporting adds.

What DataAnnotation says:

• Pay: general projects start at $25 to $50 an hour for evaluating chatbot responses, with higher rates for other tracks and for strong performance. Its home page advertises up to $125 or more an hour for most expert roles and $150 or more for coding.
• Getting paid: it pays through PayPal, and deposits arrive within a few days of requesting them.
• Flexibility: no minimum hours or daily login requirements, and projects run around the clock.
• Requirements: reliable internet and fluent English.
• Quality: it says it keeps strict quality standards and that subpar work can cost you access to future projects.

What reporting adds:

• Who’s behind it: DataAnnotation doesn’t name a parent company on its site. Time reported in April 2024 that it is reportedly a subsidiary of Surge AI, a data labeling company.
• Worker experiences: the same Time article reported that applicants who aren’t accepted typically don’t hear anything after the assessment, that some people were told they passed but were never offered tasks, and that some reported accounts deactivated with large amounts of earnings not yet paid out. DataAnnotation and Surge AI didn’t respond to Time’s request for comment.
• A lawsuit: in May 2025, Bloomberg Law reported a proposed class action in California state court against Surge Labs Inc., the company known as Surge AI, alleging that it deliberately misclassified workers as independent contractors.

What that means for you: the work and the pay are real, but approval isn’t guaranteed work, and you’re a contractor without employee protections. Request payouts regularly rather than letting a large balance build up, keep your own record of hours and earnings, and don’t rely on one platform for your whole income.

The DataAnnotation Starter Assessment

DataAnnotation’s Starter Assessment is how you get in, and it works differently from most job applications:

• You get one attempt. DataAnnotation says you can only take it once, with no retakes.
• It takes about an hour, and specialized assessments can take one to two hours.
• It’s a preview of the work. DataAnnotation says it gives you direct experience with the project types you’ll do after approval.
• You’ll hear back by email within a few days if you’re approved; Time reported that applicants who aren’t accepted typically hear nothing.

Because there’s no second chance, prepare before you start:

1. Block a quiet, uninterrupted hour or two. Don’t start it on your phone or between other tasks.
2. Read every instruction twice before answering, and check each answer against every requirement in the prompt.
3. Practice evaluating chatbot answers. Take any AI answer and check it for accuracy, whether it followed every instruction, safety, and clarity. Then write two or three sentences explaining your rating with specific evidence.
4. Check facts instead of trusting how confident an answer sounds. Fluent, well-formatted answers can still be wrong.
5. Write in your own words. DataAnnotation says it keeps strict quality standards, and platforms such as Mercor explicitly ban AI-written answers.
6. Choose the track you’re strongest in. If you have expertise in coding, a science or a profession, apply for that track rather than the general one.

How to get started with data annotation

1. Apply to two or three platforms, not just one, since acceptance and work volume vary.
2. Lead with your languages and expertise. Language skills open up search and ad evaluation roles; a degree or profession opens expert tracks.
3. Pass the qualification carefully, treating it like a one-shot exam even when retakes exist.
4. Read the project guidelines fully before your first task, and reread them when they change.
5. Track your hours and earnings yourself for taxes and in case of disputes.

Avoid data annotation scams

Real platforms don’t charge you to work. The FTC’s warnings about task scams apply directly here: an unexpected text or WhatsApp message offers money for simple online tasks, an app shows growing “earnings”, and then you’re asked to deposit money, often in crypto, to unlock more tasks or withdraw. The FTC says never to pay to get paid. CrowdGen, for example, says it will never ask you to pay to join a project. Apply only through each platform’s own website.

If you’re in the US, you’re usually paid as an independent contractor, so set money aside for self-employment tax. Our AI trainer jobs guide explains the IRS rules.

Practice for platforms that interview you

Some platforms, such as Mercor and micro1, screen applicants with an AI interview instead of a written test. Our guides to the Mercor AI interview and the micro1 AI interview cover each one.

Frequently asked questions

Sources

• DataAnnotation, home page and FAQs. Read on 8 November 2026.
• Will Henshall, Side Hustle or Scam? What to Know About Data Annotation Work, Time, 2 April 2024. Read on 8 November 2026.
• Bloomberg Law, AI Training Firm Surge AI Hit With Worker Misclassification Suit, 20 May 2025. Read on 8 November 2026.
• CrowdGen by Appen, TELUS Digital AI Community jobs, Outlier and Mindrift. Read on 8 November 2026.
• Amazon Mechanical Turk, worker page. Read on 8 November 2026.
• Federal Trade Commission, How to spot and avoid task scams. Read on 8 November 2026.