Ways to earn
Article by
Data annotation platforms connect contributors with AI companies that need people to label data, rate AI responses, and check outputs for accuracy. The main options differ in four ways that might matter more than pay: how you get in, which tasks you can access, how you get paid, and whether your country is supported.
Most lists of data annotation platforms rank sites by the highest rate they advertise. That number actually tells you very little.
A platform that pays well but doesn’t support your country, pays only through a method you can’t use, or gives you a single assessment attempt with no feedback may never pay you at all.
That’s why we’ve put together a comparison of six different platforms with annotation projects. We’ll answer questions contributors typically run into, like how the entry process works, what the tasks looks like, how money reaches you, and what changed in 2025 and 2026.
Important note: Every fact below comes from each platform’s own public pages (as of November 2026). Platforms update terms often, so treat the details as a starting point and confirm them on the sign-up page before you invest time.
Data annotation platforms compared at a glance
The table below sets the six platforms side by side on the factors that decide whether you can start and get paid.
Platform | Operated by | Entry path | Typical task types | Published payout methods | Availability notes |
|---|---|---|---|---|---|
DataAnnotation | Surge AI | Choose a track, then one Starter Assessment (about 1–2 hours), no retakes | Rating and ranking AI responses, fact-checking, prompt refinement, domain review, coding | PayPal only | Requires fluent English; country list not published |
Outlier | Scale AI | Apply to a specific track; complete skill screenings, onboarding per project | Generative AI training, evaluation, language and coding tasks | PayPal, AirTM, bank transfer | Rates vary by location; quiet periods possible |
Remotasks | Scale AI | Country-specific sign-up | Labeling, especially image and video annotation | PayPal, AirTM | Closed in several countries in 2024 |
CrowdGen | Appen | Profile, expertise verification, then a qualification per project | Evaluating, annotating, and generating data in your field; language projects | PayPal, bank transfer, AirTM, SWIFT, Payoneer, gift card via Dots, fees apply | Methods vary by country; small balances held until above fee threshold |
OneForma | Centific | Profile with languages and translation pairs; ID verification may apply | Data collection, annotation, judging, transcription, translation | Payoneer, PayPal, Tipalti | Methods vary by region |
Mindrift | Toloka | Apply, complete a domain assessment, onboard to a project | Evaluating AI responses, labeling and classifying, search relevance, plus specialized expert projects | Payoneer, Tipalti (PayPal), Tipalti (bank transfer) | Restricted-jurisdiction list published; task volume depends on open projects |
What changed in annotation since 2024
The market these platforms serve shifted quickly, and the shifts explain most of the differences in the table.
The tasks moved from drawing boxes to judging answers
Image and video labeling tasks still exist, but a large number of open projects now ask people to compare, rate, and correct responses from large language models. These projects reward reading comprehension and consistency more than speed with a labeling tool.
Crowd platforms pivoted toward experts
Previously generalist platforms like CrowdGen, formerly Appen Connect, now lead with domain projects for accountants, lawyers, clinicians, linguists, and software engineers. Generalist tasks remain, but they are no longer the headline.
Country coverage narrowed
Remotasks ended operations in Kenya, Nigeria, and Pakistan in 2024 and stopped new sign-ups in Thailand, Vietnam, and Poland. Amazon Mechanical Turk closed on 30 September 2026, which pushed many long-time task workers to look elsewhere – see what the MTurk shutdown means for task workers.
Pay ranges moved into individual listings
Several platforms replaced homepage pay ceilings with rates shown per project or per task, so the number on a landing page is a less reliable guide than it was.
Six data annotation platforms to try in 2026 (and beyond)
Each profile below covers what the platform is, how you get in, how pay works, and the one thing worth knowing before you apply.
DataAnnotation
DataAnnotation is the contributor platform associated with Surge AI. Tasks center on rating and ranking AI responses, checking facts, refining prompts, and reviewing content in specialist fields, with separate coding and language tracks.
Entry: You pick a qualification track and take one Starter Assessment. General assessments take about an hour and specialized ones one to two hours. There are no retakes.
Pay: Payments go to PayPal only, and deposits arrive within a few days of a withdrawal request. The site advertises different ceilings for generalist, language, expert, and coding tracks.
Worth knowing: The FAQ states that the team can't reply to every applicant. Silence after the assessment usually means the application is still under review or wasn't accepted, and you won't get feedback either way.
Outlier
Outlier is Scale AI's platform for generative AI training work. It runs tracks for generalists, language speakers, coders, and subject experts, and many expert tracks prefer or require a degree.
Entry: You apply to an open track and complete onboarding for each project you join.
Pay: Payouts go through PayPal, AirTM, or bank transfer and are processed weekly. The FAQ gives a broad hourly range that depends on skill level, task type, location, and project, and rates are shown before you accept a task.
Worth knowing: Outlier states there are no minimum hours, and also that periods without available work happen because volume depends on customer demand. They also publish a list of allowed countries to contribute from.
Remotasks
Remotasks is Scale AI's older labeling platform, set up in 2017 with a focus on computer vision data for autonomous vehicles. It is still operating under terms updated in October 2025.
Entry: Sign-up depends on your country, so check the registration page for your location before anything else.
Pay: PayPal and AirTM are the two payout options.
Worth knowing: In 2024 Remotasks closed in Kenya, Nigeria, and Pakistan and stopped accepting new sign-ups in Thailand, Vietnam, and Poland. If you are in an affected country, the platform is no longer an accessible option.
CrowdGen
CrowdGen is Appen's contributor community, renamed from Appen Connect. Appen has run crowd projects since 1996 and states it has paid more than 500,000 contributors.
Entry: You create a profile, your expertise is verified, and you are matched to projects in your field. Most projects then have their own qualification step.
Pay: CrowdGen offers payout options like PayPal, local bank transfer, AirTM, SWIFT, Payoneer, and gift cards. Appen says it charges no fees itself, but third-party fees apply and differ by method and country.
Worth knowing: If a payment is smaller than the fees it would trigger, CrowdGen holds it until a later month. Small, occasional earnings can therefore arrive later than expected.
OneForma
OneForma is Centific's freelance platform, with a strong focus on language work. Its project categories are data collection, annotation, judging, transcription, and translation.
Entry: You build a profile that lists your native and spoken languages and translation pairs, which drives project matching. Identity verification may be required for some projects.
Pay: The help center supports Payoneer, PayPal, and Tipalti, with availability by region, and notes a yearly limit on PayPal payouts.
Worth knowing: Language skills are the main lever here. Bilingual contributors usually see more matching projects than English-only ones.
Mindrift
Mindrift is an AI training platform run by Toloka, operating in AI data since 2014. Our AI data evaluation and annotation projects focus on evaluating AI responses, labeling and classifying content, and judging search relevance, alongside specialized projects for experts in fields like coding, STEM, and law.
Entry: You apply, complete an assessment matched to the project, and then onboard with project guidelines. The guide to how Mindrift assessments work explains what the tests check.
Pay: Rates are set per task and shown before you start. Evaluation projects currently list hourly-equivalent rates between $5 and $30, depending on task complexity. Payouts go through Payoneer or Tipalti (PayPal) depending on your country and are processed twice a month, on the 5th and the 20th. Details are in how Mindrift pays contributors.
Worth knowing: Task volume depends on which client projects are open, so quiet periods can happen. Contributors in countries without Payoneer or Tipalti support can't join, and the full eligibility and geographic restrictions list is worth checking before you apply.
How to choose a platform for your profile
The right platform depends less on advertised pay and more on which of these profiles fits you.
If you are… | Start with | Why |
|---|---|---|
A careful reader with no degree or specialist background | Platforms with generalist or evaluation tracks: Mindrift, DataAnnotation, Outlier | Response rating rewards judgment and guideline-following, not credentials |
Bilingual or fluent in a less common language | Mindrift, OneForma, CrowdGen, Outlier language tracks | Language matching opens projects that English-only contributors can't take |
A domain expert (law, medicine, finance, code) | Mindrift specialized projects, CrowdGen, Outlier expert tracks | Expert projects carry higher rates and fewer competing contributors |
Outside the US, UK, or Canada | Platforms with wide payout options: Mindrift, CrowdGen, OneForma | Payout method support decides eligibility more often than skill does |
If more than one row fits, start where your strongest skill is rewarded and add a second platform once you understand how tasks and reviews work.
Where contributors get stuck
Public contributor forums repeat the same five frustrations across every platform in this list. Knowing them in advance sets realistic expectations and prepares you in advance.
Silence after applying: Several platforms don't reply to unsuccessful applicants. Set a date to move on rather than waiting indefinitely.
Single-attempt assessments: Read the guidelines twice and take the test when you have uninterrupted time. Speed rarely matters as much as accuracy.
Quiet periods: Task volume follows client demand on every platform. On Mindrift, how task availability works explains why a quiet dashboard usually means no open project matches your qualifications yet.
Payout friction: Fees, minimum thresholds, and held balances differ by method. Check the fee for your country before choosing a payout option.
Rejected work: Tasks that fail quality review usually aren't paid. Consistency with the guidelines matters more than volume.
None of these mean a platform is a scam. They are normal features of project-based opportunities, and they account for most negative reviews.
Red flags no real annotation platform shows
Legitimate data annotation companies pay contributors — never the other way around. Walk away from any offer that asks for a registration fee, paid training or certification, a deposit before your first payout, or contact only through private messaging apps.
The FTC's guidance on job scams covers the same warning signs, and our checklist for spotting fake AI training platforms applies them to this niche. Impersonation is the more common risk. Scammers copy the names of well-known platforms, so always apply through the platform's own website rather than a link sent in a message.
Running more than one platform
Many contributors join two or three platforms to smooth out quiet periods. A few habits make that manageable:
Stagger assessments: Take one at a time so each gets your full attention.
Track payouts by platform: Payout schedules range from weekly to monthly, so a simple spreadsheet prevents missed or held payments.
Read each platform's terms: Rules on confidentiality and conflicts differ, and project guidelines are usually confidential.
Keep tax records from day one: Contributors on these platforms are typically independent and handle their own tax reporting. US contributors can start with the IRS gig economy tax center.
A small amount of admin up front is what keeps multiple income streams from turning into missed earnings.
Article by
Mindrift Team
