Is DataAnnotation Legit? Its Own Site Quotes 3 Pay Ranges

DataAnnotation can’t agree with itself about money. Its FAQ says it has paid contractors “over $20 million” since 2020. Its About page says $150M+. On the homepage, the headline offers expert rates of up to $75 to $125+ an hour, the description search engines show says $50 to $150+, and the blurb that appears when you share the link says $20 to $60+. All five figures were live on the site when we checked on October 1, 2026.

It looks like stale marketing copy rather than fraud. DataAnnotation is a real platform run by Surge AI, one of the largest AI data-labeling companies, and workers do get paid. Our verdict is Caution. The money is real, but the pay most people land on, the one-shot unpaid entry test, accounts that go quiet without explanation and a privacy clause about selling task data all deserve a look before you apply.

Last checked: October 1, 2026. How we checked this is explained at the end.

Who’s actually behind it

For years the site didn’t say. Surge AI’s Wikipedia entry notes that the platform was criticized for not being clear about who owns it, and lists DataAnnotation as a subsidiary of Surge Labs Inc. Surge was founded by Edwin Chen in 2020 and counts Anthropic, OpenAI, Microsoft and Meta among its customers. The homepage logo now reads “DataAnnotation by Surge,” though the About page and FAQ still don’t name the company.

The clearest confirmation comes from a court filing. In May 2025, Clarkson Law Firm filed a proposed class action in San Francisco Superior Court against “Surge Labs, Inc. dba DataAnnotation” and Chen personally (Cavalier v. Surge Labs, case CGC-25-625502). It alleges annotators are misclassified as independent contractors, aren’t paid for required training time, and face strict task time limits that cut their pay.

Nothing has been proven. We couldn’t find a ruling or settlement in public records as of October 1, 2026, so treat the case as open.

What the job board actually pays

The homepage lists 107 roles, each with a pay range and a “hired recently” count. We copied every one on October 1 and grouped them. The hiring counts are DataAnnotation’s own and we can’t verify them, but they’re the best picture the company gives of where new workers land.

Here’s the arithmetic the ads skip. Of the 6,188 recent hires the board shows, 2,819 (46%) went into language roles that top out at $40 an hour. Add the single generalist role, capped at $50, and 53% of recent hires sit in roles where the ceiling is $50 or less. The $125 and $150 figures belong to roles that want a law license, a medical degree, a PhD or real coding skill.

Worker reports land lower still. Breaking Even, a site that tracks AI gig platforms, says generalists typically report $20 to $25 an hour and coders $40 to $45. Even the FAQ wobbles: one answer puts general projects at $25 to $50+, another at $25 to $30+.

DataAnnotation’s homepage and job board, screenshot taken October 1, 2026. The “38” next to Languages is the largest category on the board.

What it costs to get in

There’s no sign-up fee. The company’s Trust and Safety page says it will never ask you for money of any kind.

The real cost is the Starter Assessment. The FAQ says most take about an hour and specialist versions one to two. You get exactly one attempt, with “no retakes or second chances,” in the company’s words. Workers say it isn’t paid, and failing usually means silence rather than a rejection email. Breaking Even’s rule of thumb: if you’ve heard nothing after two or three weeks, read it as a no.

Support is thin by design. The FAQ admits the company can’t respond to every applicant, and says weak work can cost you access to future projects.

Where the complaints pile up

On Trustpilot, DataAnnotation scores 3.9 from 1,969 reviews: 55% five-star, 20% one-star and very little in between. Trustpilot also notes the company hasn’t invited reviews recently and hasn’t replied to negative ones. So the score isn’t padded by invitations, but complaints sit there unanswered.

Breaking Even scored 2,087 Reddit discussions about the platform from May 29 to August 17, 2026. By its count, 34% were negative, 55% neutral and 11% positive. Empty queues came up in 484 threads, no work available in 383 and technical problems in 241. High earnings came up in 127.

Recent one-star reviews, all second-hand, describe:

• Applications and finished assessments that got no reply for weeks or months.
• A growing list of unpaid qualification tests just to unlock individual projects, from a reviewer who’d worked there for over a year.
• An account locked “under review” with a four-figure balance while the worker was asked to justify old tasks.
• Repeated identity re-verification requests and no answer from support.

The five-star reviews are just as consistent: on-time payments, interesting work, real flexibility. Both camps are describing the same machine. When work flows, it pays. When it stops, nobody tells you why.

The privacy line most applicants skip

DataAnnotation’s privacy notice, effective June 23, 2026, has two sentences worth reading twice. It says data collected through the website isn’t sold, but that if you take part in a specific task, “your data may be sold” under that task’s agreement. It also says that once your data reaches a third party as part of a task, the company can’t retrieve or delete it.

That matters most for tasks involving your voice, face or personal details. The same notice says contractors may have facial recognition data or other biometric identifiers collected, and the FAQ says ID and biometric data from verification are stored by both DataAnnotation and Persona, its identity-check provider.

Four habits keep you in control:

1. Read the agreement for any project that records audio, video or personal details before you accept it. That’s where any sale of your data is spelled out.
2. Save a copy of every agreement you accept.
3. To close your account and ask for deletion, email [email protected]. The notice says data is generally kept for two years after your last contact, and longer where tax or legal rules require it.
4. Assume anything already delivered to a client stays with that client.

Is that message really from DataAnnotation?

A well-known name is a gift to scammers. DataAnnotation’s own footer says “No recruiters. No interviews.” Its Trust and Safety page says the only way to apply is the official sign-up flow at app.dataannotation.tech, and tells you not to engage with anyone contacting you outside its official channels. Use that as your filter:

• A LinkedIn, WhatsApp, Telegram or text “recruiter” offering you DataAnnotation work is a red flag. Apply yourself, from the real site.
• A request to deposit money, buy crypto or pay a fee to unlock tasks or withdraw earnings is a task scam, the exact pattern the FTC describes.
• A sign-up link on any domain other than dataannotation.tech isn’t the real thing.
• Anyone asking for bank or card details while you’re not logged in isn’t DataAnnotation. The company says it only asks inside the platform, and only to pay you.

One August Trustpilot reviewer described being asked to pay through a registration process and then being blocked. That matches impersonation scams, since the real platform says it never asks for money. If you’ve paid someone, report it at ReportFraud.ftc.gov and contact your bank or PayPal straight away.

So should you apply?

Apply if you have a credential the board pays for, a second language it’s hiring in, or strong writing skills, and you want side income you can pick up and put down. Skip it if you need predictable hours, or if you aren’t comfortable with ID checks and task data that may be passed to clients.

It’s far safer than any platform that charges you to start, and it belongs to the same family as the others we’ve reviewed. Outlier AI pays weekly but has the same empty-queue problem. Alignerr advertises an $80/hr average its own listings don’t match. Mercor pays specialists well but leaked contractor data in 2026, and Handshake AI only takes U.S.-based fellows. Running two or three at once is the steadiest setup.

FAQs