AI jobs

DataAnnotation Review

What DataAnnotation actually pays, the one-attempt Starter Assessment that decides everything, who it accepts, and the work-availability problem that defines the 2026 experience.

Updated August 2026

Quick answer: DataAnnotation is a real platform that pays real contributors, and it is one of the better-paying entry points in AI training work. Its own FAQ advertises $25–30+/hr for general projects and $50–100+/hr for coding, STEM and professional tracks, paid per hour logged via PayPal. Two things decide your experience, and neither is whether you get paid. First, the Starter Assessment is unpaid, takes about an hour, and you get exactly one attempt — no retakes, no feedback. Second, work availability is inconsistent, and 2026 contributor reviews are dominated by dry spells rather than pay complaints. It accepts contributors from six countries only. Treat it as variable supplemental income, not a job.

For the full platform landscape, see get paid to train AI. For how it compares with the other major platform, see our Outlier review.

Who actually owns DataAnnotation

Worth knowing, because the worker-facing site does not say it: DataAnnotation is operated by Surge Labs, Inc., trading as Surge AI, founded in 2020 by Edwin Chen, who holds roughly a 75% stake. Surge reported $1.2 billion in revenue in 2024 — ahead of Scale AI’s $870 million that year — with customers including Google, Meta, Microsoft, Anthropic and Mistral, and it has taken no outside funding (Forbes, 17 September 2025).

That matters for one practical reason: it tells you the platform is backed by a large, revenue-generating parent rather than a thinly capitalised startup, which is a reasonable thing to want to know before investing unpaid hours in an assessment.

What the work actually is

You are shown something an AI model produced and you judge it, or you produce the material the model should have produced. In practice:

• Generalist work — evaluating chatbot responses, comparing outputs from different models, testing image generation, writing and editing prompts.
• Coding — code evaluation and debugging across Python, JavaScript, HTML, C++, C# and SQL.
• STEM — mathematics, physics, chemistry and biology projects.
• Professional tracks — law, finance, accounting and medicine, gated behind credential verification.
• Bilingual and multilingual — translation, localisation and cross-language annotation.

Work appears on a dashboard and you choose from the projects you have qualified for. Access to specific projects opens up as you demonstrate quality — DataAnnotation’s FAQ states that workers are added to projects based on expertise and performance, and that access can be withdrawn for poor-quality contributions.

The Starter Assessment: the thing that decides everything

This is the single most searched aspect of the platform, and the most important operational fact on this page.

You get one attempt. DataAnnotation’s own FAQ states plainly that the Starter Assessment can only be taken once, with no retakes and no second chances. Most people treat it as a form to rush through. It is the interview.

DataAnnotation says it reviews submissions for thoroughness and accuracy rather than speed. Its own blog cites a 2.6% acceptance rate — a company claim, unaudited and unverifiable, but consistent with how selective contributors report the process being.

How to actually prepare. Read the entire brief and rubric before writing anything. Answer at length rather than efficiently. Proofread. The most commonly reported cause of rejection is skimming the instructions, not lacking the ability. Budget two to three hours even though the estimate says one, and do it when you are fresh rather than at the end of a working day.

What it pays

DataAnnotation publishes rate ranges by track, which is more transparency than most of the category offers.

Why the advertised and effective figures differ so much, and none of it is hidden:

You are paid per hour logged, not per task. You run a timer in the platform while working. The advertised rate is the rate while you are actively working on an approved project.

The assessment and qualification tests are unpaid. So is time spent checking whether any projects are available.

Dry spells are the real rate-killer. An advertised $30/hr becomes something quite different if you can only log four hours in a month. This is the single biggest gap between the headline number and what lands in your PayPal account.

The practical test: track your own effective rate across a full month, counting the unpaid assessment time, before deciding whether it is worth continuing.

Who can join

Countries. DataAnnotation does not publish an eligibility list on its FAQ or homepage. Third-party reviews consistently report six — the United States, Canada, the United Kingdom, Ireland, Australia and New Zealand — but that figure traces only to those reviews, so treat it as unconfirmed and check at signup. What the company does state is that you can create an account from elsewhere and be notified if eligibility expands.

Credentials. The stated baseline is a bachelor’s degree or equivalent real-world experience for generalist work. Specialist tracks require the relevant advanced degree or professional credential.

Language and equipment. Fluent English and reliable internet. A smartphone is useful for the identity verification step.

Identity verification. Handled by Persona, a third-party provider. You will need a government-issued photo ID, front and back, plus a front-facing selfie. Turn any VPN off before starting, as it commonly causes verification to fail.

Tax status. Independent contractor — 1099 in the United States, self-employment obligations elsewhere. No benefits, no guaranteed hours.

How you get paid

Note this is not a weekly payroll. Money becomes available on a rolling basis as each batch of work clears its approval window, and you request withdrawals yourself. If you have no PayPal account, you have no way to be paid.

Standing advice for any gig platform: withdraw as your balance clears rather than letting it build up.

Work availability: the defining issue in 2026

If you read one section before signing up, make it this one.

Across Trustpilot, Glassdoor, Indeed and Reddit, the dominant 2026 theme is not pay and not payment reliability — it is that there is often nothing to work on. Contributors describe weeks with an empty dashboard, and months averaging only a handful of billable hours. The pattern is notably concentrated in 2026 relative to earlier years.

This is a category-wide shift rather than something specific to DataAnnotation. Generalist annotation volume is contracting while credentialed specialist work holds up. In September 2025, xAI cut around 500 of its roughly 1,500 data annotators, telling staff it would scale back generalist roles and expand specialist AI tutors tenfold across STEM, finance, medicine and safety (Business Insider, 12 September 2025).

What this means practically: do not plan around DataAnnotation as a single income source. The contributors reporting the best results hold qualifications on two or three platforms and work whichever has volume that week. If you hold a coding, STEM, legal, financial or medical credential, your position is considerably stronger than a generalist’s.

Contributor ratings

All figures observed August 2026. They move, so treat them as a snapshot.

Glassdoor records 88% of reviewers saying they would recommend it, with work-life balance at 4.7 and compensation at 3.9, but career opportunities at 3.4. Indeed shows the same shape: work-life balance 4.6, pay 4.1, job security and advancement 3.1.

Read that pattern rather than the headline number. Contributors rate the flexibility and the pay rate highly and rate the security and progression poorly — which is an accurate description of well-paid freelance work with no guaranteed volume. It scores well as a side income and badly as a job, because that is what it is.

One thing to check when reading reviews elsewhere: there are separate Trustpilot listings for app.dataannotation.tech and dataannotation.com with tiny review counts and lower scores. The main platform listing is dataannotation.tech.

Is it a scam? No — but real scams use its name

DataAnnotation is a legitimate operation: a substantial parent company, published rates, third-party identity verification, PayPal payment, and thousands of corroborating contributor reviews. It never asks anyone for money, and its FAQ says so explicitly.

The scams are impersonators, and they are common enough to be worth naming:

• Fake recruiters contacting people by text, WhatsApp or Telegram. DataAnnotation contacts applicants by email after the assessment.
• Lookalike domains — variants of the real address with different endings. The real platform is dataannotation.tech.
• Any request for payment for training, equipment, software or access. No legitimate platform in this category charges you anything.
• Cheque-deposit schemes, where you are asked to deposit a cheque and return part of the funds.

Why legitimate applicants conclude it is a scam. Three experiences drive most of it, and none involves fraud: being rejected at the assessment with no explanation and no retake; being accepted and then finding no available projects for weeks; and losing access to a project after a quality score drops, which the FAQ says can happen. All three are frustrating. None of them means the platform is not real.

Who it suits, and who should skip it

Worth your time if:

• You hold a credential in coding, STEM, law, finance or medicine. The rates are among the best in the category and the specialist queues are where the remaining volume sits.
• You live in one of the six eligible countries and can afford to spend two or three careful unpaid hours on a one-attempt assessment.
• You want flexible supplemental income and can tolerate months where very little is available.

Look elsewhere if:

• You need predictable weekly income.
• You are outside the six eligible countries — create an account for the waiting list, but do not plan around it.
• You have no PayPal account and cannot open one.
• You want to rush the assessment. With one attempt and no feedback, a hurried submission ends the opportunity permanently.

How DataAnnotation compares

The short version: DataAnnotation pays better and screens harder; Outlier is easier to get into and has more work available; Appen is the fallback if you are outside the wealthy-country list. Running two of them is more sensible than committing to one. Full detail on get paid to train AI.

Frequently asked questions

Is DataAnnotation legit?

Yes. DataAnnotation is operated by Surge Labs, trading as Surge AI, a company that reported $1.2 billion in revenue in 2024, and it pays contributors via PayPal with no fees charged at any point. It has thousands of reviews averaging around 4 out of 5 across Trustpilot, Indeed and Glassdoor. Scams do use its name — fake recruiters on WhatsApp or Telegram and lookalike domains — so apply only at dataannotation.tech, and treat any request for money as fraudulent.

How much does DataAnnotation pay?

Its published rates are $25–30+/hr for general projects, $20+/hr for multilingual work, and $50–100+/hr for coding, STEM and professional tracks. Those are the rates while you are actively logged into an approved project. Contributor-reported effective earnings for generalist work are commonly nearer $14–20/hr once you account for the unpaid assessment and periods without available work, and Glassdoor’s aggregate of contributor submissions estimates about $39/hr.

How long does the DataAnnotation assessment take?

DataAnnotation states most Starter Assessments take about an hour, with specialist versions in coding, mathematics, sciences, law, finance, medicine and specific languages taking one to two hours. Contributors commonly report spending two to three hours because the assessment rewards thoroughness. It is unpaid.

Can you retake the DataAnnotation assessment?

No. DataAnnotation’s FAQ states the Starter Assessment can only be taken once, with no retakes and no second chances. You will not receive feedback explaining a rejection. This is the reason to read the full brief before starting and to proofread before submitting — the most commonly reported cause of failure is rushing rather than lack of ability.

Why are there no projects on DataAnnotation?

Work availability depends on client demand and on which projects you have qualified for, and 2026 has seen widespread reports of dry spells. The wider pattern across the industry is a shift away from generalist annotation toward credentialed specialist work, so generalist queues have thinned while coding, STEM and professional tracks have held up better. Holding qualifications on more than one platform is the standard response.

Does DataAnnotation pay weekly?

Not on a fixed weekly cycle. Task pay is approved around seven days after you submit the work, and once approved you can withdraw to PayPal roughly every three days at your own discretion. There is no published minimum payout, and PayPal is the only payment method.

Who owns DataAnnotation?

DataAnnotation is operated by Surge Labs, Inc., trading as Surge AI, founded in 2020 by Edwin Chen, who holds roughly a 75% stake. Surge’s clients include Google, Meta, Microsoft, Anthropic and Mistral. The worker-facing site does not mention Surge by name, which is why the question comes up so often.

Which countries does DataAnnotation accept?

Third-party reviews consistently report six — the United States, Canada, the United Kingdom, Ireland, Australia and New Zealand — but DataAnnotation publishes no eligibility list on its own site, so that figure traces only to those reviews. Treat it as unconfirmed and check at signup. If you are outside those countries you can still register an account, and the company says it will contact you if eligibility expands.

Is DataAnnotation worth it?

For credentialed specialists in the six eligible countries, generally yes — the published rates are among the best in the category. For generalists the case is weaker than it was in 2024, because the assessment is selective, there is one attempt, and generalist project volume has thinned. Judge it on your own effective hourly rate across a full month rather than on the advertised rate, and treat it as supplemental income rather than a replacement for a job.

Related pages

Start here: AI jobs hub · Best AI jobs for beginners · AI salaries

Platform reviews: Outlier review

The work itself: Get paid to train AI · AI training jobs · Data annotation jobs

Other routes: Remote AI jobs · AI side hustles · AI career paths

Platform terms, advertised rates and contributor ratings were checked against DataAnnotation’s own published pages and third-party review platforms on 17 August 2026, and change without notice. Company figures including advertised rates, total contributor payments and acceptance rate are unaudited company claims, labelled as such in the text. There is no independent audited survey of pay in this category, so every contributor-reported figure is a self-reported estimate.