How to pass the DataAnnotation Starter Assessment

DataAnnotation doesn't interview you. It tests you once. The Starter Assessment is the whole application: no CV review, no call, and no second try. If it goes well you get an email within days. If it doesn't, you usually hear nothing at all. Here's what the assessment is, what it tests, and how to prepare without falling for "answer" sellers.

We haven't taken the assessment ourselves. This guide is researched from DataAnnotation's own FAQ and role pages, checked on 7 October 2026, plus a court filing. Where DataAnnotation doesn't publish something, we say so.

What the DataAnnotation assessment actually is

When you sign up on dataannotation.tech you pick a track, and each track has its own Starter Assessment. The FAQ says you choose "from specialized Starter Assessments" so your test matches what you can actually do.

The tracks, from DataAnnotation's FAQ:

• General: rating chatbot answers, comparing two outputs, checking facts, testing image generation.
• Multilingual: translation and localisation. You need native fluency in a language beyond English.
• Coding: reviewing and debugging AI-written code.
• STEM: maths, physics, biology or chemistry, for people with a master's or PhD, or a bachelor's plus 10 years' experience.
• Professional: law, finance or medicine, for people with licensed credentials.

After you pass, you can take more specialist assessments to unlock higher-paying projects. The FAQ names coding, maths, chemistry, biology, physics, finance, law, medicine and language-specific tests.

Starter, Core or specialist?

You'll see "Core Assessment" in forums and third-party guides. DataAnnotation's own FAQ doesn't use that name. It describes one Starter Assessment per track, then optional specialist assessments once you're in. We couldn't confirm how a "Core" test works from anything DataAnnotation publishes, so treat any blog that explains it in detail with care.

How long it takes

• Starter Assessment: "Most Starter Assessments take about an hour," the FAQ says.
• Specialist assessments: one to two hours, depending on the subject.
• Different pages, different numbers: DataAnnotation's Generalist page says "~30 min assessment" and its Coding page says "~60 min assessment".

There's no published time limit, and the FAQ says reviewers judge "thoroughness and accuracy rather than speed". Block out two hours anyway. DataAnnotation doesn't say whether the assessment is paid. A 2025 lawsuit alleges onboarding assessments are unpaid (more below), so plan it as unpaid time.

What it tests

DataAnnotation says the Starter Assessment "gives you direct experience with project types you'll work on after approval". So it's a sample of the real work, not a quiz about AI.

What each track looks for, in DataAnnotation's words:

Track	What it says it checks
General	Strong English, critical thinking, research and fact-checking. Its Generalist page adds spotting "logical flaws", catching "factual errors" and writing "clear feedback".
Coding	Python, JavaScript or another language, plus the "ability to solve LeetCode-style problems". Its Coding page adds finding "correctness bugs", reasoning about complexity and writing "clean, idiomatic code".
Multilingual	Native fluency in more than one language on top of English.
STEM and Professional	Advanced knowledge in the field, backed by a degree or licence.

Every track also needs "self-motivation and ability to follow detailed instructions independently". That last part is where most people go wrong.

On degrees, the site disagrees with itself. The FAQ says the baseline for general work is "a bachelor's degree or equivalent real-world experience". The Generalist role page on its job board says "No degree required." Pick the track that fits what you can do, and don't stretch.

The coding assessment

If you pick coding, expect to read and judge code as much as write it. The FAQ lists Python, JavaScript, HTML, C++, C# and SQL as languages its coding projects use. The Software Engineer page on its job board says most tasks use Python, C++, Rust, Go or JavaScript. Choose the language you know best, not the one that sounds most impressive.

What it pays once you're in

DataAnnotation's FAQ lists starting rates by track (checked 7 October 2026):

Track	Starting rate
General	$25–$50+/hr
Multilingual	$20+/hr
Coding	$40–$150+/hr
STEM	$40–$125+/hr
Professional	$40–$125+/hr

The same FAQ also says general projects "start at $25-$30+ per hour" and coding projects "range from $50 to $100 per hour", so read the table as a range, not a promise. Pay goes out by PayPal, "within a few days" of your request. More on pay and the company is in our DataAnnotation review.

How to prepare

There's no answer key to study, and anyone selling one is either guessing or scamming. What you can do is prepare to work carefully for an hour or two.

Setup checklist

• Apply only through the sign-up flow on dataannotation.tech. Its Trust & Safety page says that's "the only way to apply", and that it will never ask you for money.
• Use a computer, not a phone. You'll be reading long instructions and writing full explanations.
• Block out two quiet hours. Don't start 20 minutes before you have to leave.
• Choose your track honestly. You can't retake it.
• Have a PayPal account and a valid government photo ID ready for later. You'll need both if you pass.

How to answer

• Read every instruction twice. Then do exactly what it says. If it asks for two sentences, don't write one or five.
• Explain your reasoning in full. The FAQ says "successful applicants spend more time crafting thorough answers rather than rushing through responses". Say why one answer is better, not just which.
• Check facts, don't guess. If a chatbot answer gives a date, a number or a name, look it up before you judge it.
• Be specific about mistakes. "Response B is wrong" is weak. "Response B says the treaty was signed in 1921; it was 1919" is what reviewers look for.
• Write it yourself. The work you're being tested on is judging AI writing. Pasting a chatbot's answer in is the one thing the assessment is built to catch, and it's dishonest.
• Proofread before you submit. Spelling and grammar are part of "strong English".

For the coding track

• Explain bugs in plain English: what breaks, on which input, and why.
• Say what the time and space complexity is when it matters, and whether a better approach exists.
• Test edge cases in your head: empty input, one item, duplicates, very large values.
• Write code you'd be happy to have reviewed. Clear names beat clever one-liners.

What not to do

• Don't buy answers or a "guaranteed pass". Nobody outside DataAnnotation knows what your test contains, and paying a stranger is how scams start. Our scam guide covers the patterns.
• Don't apply through a link someone sends you. DataAnnotation says it doesn't process applications from any other route.
• Don't rush to finish first. Speed isn't scored. Care is.
• Don't open a second account to try again. The FAQ says there are "no retakes or second chances", and you'll verify your government ID with Persona before you can work.

After you submit: passing, or never hearing back

If you pass, DataAnnotation says it emails you "within a few days" and you "gain immediate platform access and can begin selecting projects". You'll then be asked to verify your identity through Persona: photos of a physical, unexpired government photo ID, then a selfie where you turn your head from side to side. The FAQ asks you to switch off any VPN first.

If you hear nothing, the FAQ says your application is "likely" still under review or hasn't been accepted. It also says it is "unable to provide direct responses for every applicant". DataAnnotation doesn't publish a timeline beyond "a few days". If a few weeks pass with no email, our advice is to plan as though the answer is no, and keep checking your inbox and spam folder.

There is no retake. The FAQ says you can take the Starter Assessment only once. The specialist assessments you can take after passing are a separate thing. They don't give you a second shot at the first test.

Passing isn't the finish line. The FAQ says work depends on demand for your skills and your past performance, and that "subpar or problematic contributions may result in loss of access to future projects". Treat your first tasks with the same care as the assessment.

What to do while you wait

Don't put your search on hold for one email that may never come.

• Apply somewhere else this week. The job list shows remote AI training roles we've checked in the last ten days, and the tracker shows which platforms are hiring hardest.
• Compare your options. Our guide to Outlier AI alternatives lays out eight platforms side by side, including DataAnnotation, with how you get in and what they pay.
• If you pass but get no projects, read passed the interview but no projects.

The catch

• One try, probably unpaid. You do an hour or two of careful work, once, with no feedback, and DataAnnotation doesn't say it pays for it.
• Silence is the usual no. You may never learn whether you failed or were simply not needed.
• There's a lawsuit about unpaid time. A May 2025 class action in San Francisco names "Surge Labs, Inc. dba DataAnnotation" (better known as Surge AI). It alleges workers were misclassified as contractors and had to complete "onboarding assessments and qualification exams, for which they are not paid". These are allegations, not findings.

DataAnnotation is still one of the better-paying places to start if you have strong English or a skill it wants. Give the assessment your full attention, then get on with other applications.

Sources we checked (7 Oct 2026)

• DataAnnotation, FAQ (assessments, review times, pay, tracks, ID verification)
• DataAnnotation, Generalist page and Coding page (assessment lengths and what they check)
• DataAnnotation job board: Generalist and Software Engineer
• DataAnnotation, Trust & Safety
• Superior Court of California, San Francisco, class action complaint against Surge Labs, Inc. dba DataAnnotation (20 May 2025)
• Clarkson Law Firm, Clarkson represents Surge AI workers in labor law class action (20 May 2025)