DataAnnotation vs Appen (CrowdGen): Pay, Tasks and Who Can Join

Both platforms sit on the human side of model training: contributors read AI outputs, rate them against guidelines, compare responses and write up their reasoning, and that feedback becomes training signal through methods such as RLHF (reinforcement learning from human feedback). Appen describes its business more broadly as "AI training data, data annotation and model evaluation for the world's leading AI teams" (appen.com, read 21 October 2026), so its projects are not limited to language-model work. The two differ most in access and mechanics, which is what this page compares, with a source and a read date for every fact. It does not tell you which one "pays more", because the two companies publish different kinds of pay figure and the figures are not comparable.

How this page is sourced. Appen and CrowdGen facts come from Appen's and CrowdGen's own pages, read on the dates shown. DataAnnotation facts are limited to what our evidence-based review establishes from 7,514 Reddit comments and DataAnnotation's own published figures, and nothing is claimed about DataAnnotation here that the review does not say. Our methodology explains the evidence bar.

What are you really choosing between with DataAnnotation vs Appen?

DataAnnotation.tech is a single platform where approved workers pick projects from a dashboard rather than being assigned to them (our review, workers' reports: 2 accounts across 2 threads, 2024 to 2026). The work is centred on language-model evaluation: rating responses, comparing them, writing justifications against a rubric. The company publishes starting rates by project type rather than a rate card, and the most reported problem in the Reddit record is not pay but supply: long stretches with nothing on the dashboard (10 accounts across 6 threads, 2026) and access ending with no explanation (8 accounts across 5 threads, 2026).

Appen is a publicly listed company (ASX code APX, founded 1996, per crowdgen.com, read 21 October 2026) that supplies training data to AI developers, and CrowdGen is its contributor platform, the renamed successor to Appen Connect. Contributors apply to individual project openings, complete a profile (age, gender, location, languages, skills) and are matched to projects; "most of our projects have a qualification process", in the platform's own words. Appen says it works "across 170+ countries and 100+ languages" (appen.com/careers, read 21 October 2026), and CrowdGen's payment system supports contributors "from over 200 countries" (crowdgen.com/crowdgen-by-appen, read 21 October 2026).

So the choice is less "which is better" than "which one you can join, and how you want work and pay to arrive": a dashboard you pick from, paid weekly-ish through PayPal, open in six countries; or project-by-project matching, paid monthly, open almost everywhere.

Facts at a glance (with read dates)

Appen is now CrowdGen: what changed for contributors

"Appen Connect is now CrowdGen by Appen" (crowdgen.com/crowdgen-by-appen, read 21 October 2026). Existing contributors sign in with their Appen credentials; new contributors apply for a project opening and an account is created for them. The platform matches you to projects from your profile, lets you set your availability, and has a payments system that, by Appen's own description, supports more countries and more payout methods than before. One practical consequence from CrowdGen's payments page (read 21 October 2026): "all existing Appen independent contractors must set up a payment method on the CrowdGen platform, even if you have previously done so on Appen Connect." If you search for "Appen Connect" today, CrowdGen is where it went.

Which pays more, DataAnnotation or Appen? What each actually publishes

The honest answer is that the two publish different things and the figures cannot be ranked against each other.

• DataAnnotation publishes floors. Its site states starting rates for different kinds of work rather than a fixed rate card: $20, $25, $30, $50 and $100 per hour depending on the type of project, as read on 15 September 2026 for our review. They are starting points the company itself states: not earnings, not an average, and not a promise that work exists.
• CrowdGen publishes ceilings. Its project cards give "up to" rates per domain and no starting rate: for example "Up to $95/hr" for linguists, "Up to $150/hr" for software engineers and "Up to $450/hr" for medical clinicians (crowdgen.com, read 21 October 2026). An "up to" figure is the top of a range for a credentialed specialist, not what a generalist task pays, and it is Appen's own published statement, not an earnings figure.

If you need one answer: for a generalist in one of DataAnnotation's six countries, DataAnnotation is the only one of the two that tells you a minimum before you start ($20 per hour, its own published starting rate), while CrowdGen publishes no rate at all for generalist work and its ceilings apply to credentialed specialists. For a specialist with a licence or a degree, compare the specific CrowdGen project ceiling with DataAnnotation's $50 and $100 starting tiers for your field. A floor and a ceiling are not otherwise comparable, so this page prints no blanket "which pays more" verdict, and you should be suspicious of pages that do: the figures circulating in AI-generated summaries of this comparison trace back to videos and forum posts, not to either company. What does differ, and is documented, is timing: DataAnnotation hours become withdrawable about a week after reporting, while CrowdGen pays monthly by the 14th for the previous month.

For the platforms that publish hourly ranges on public listings, our daily rate dataset tracks the advertised medians with dates. Appen is not in that dataset (it is not one of the platforms whose listings carry comparable hourly figures), and DataAnnotation posts no public listings, so neither appears there.

Does geographic location limit your options?

Yes, and it is the first filter. Workers describe DataAnnotation as open in six countries (United States, Canada, United Kingdom, New Zealand, Australia and Ireland), with individual projects narrowing eligibility further (our review, 4 accounts across 4 threads, 2023 to 2026). If you are outside those six, DataAnnotation is not an option regardless of your skills.

Appen describes itself as working "across 170+ countries and 100+ languages" (appen.com/careers, read 21 October 2026), and CrowdGen's payments system supports contributors "from over 200 countries" (crowdgen.com/crowdgen-by-appen, read 21 October 2026). Project availability still depends on your location and languages, because projects are matched to profiles, but the door is open almost everywhere.

How consistent is the work?

DataAnnotation. The most reported problem in the review's 7,514-comment record is supply, not pay: long stretches with nothing on the dashboard (10 accounts across 6 threads, 2026), tasks that simply stop appearing (8 accounts across 6 threads, 2024 to 2025), and access ending with no explanation, sometimes shortly after positive feedback (8 accounts across 5 threads, 2026). One avoidable trigger is documented: connecting through a VPN, or a shared connection that resolves to one, can get an account flagged (2 accounts across 2 threads, 2026). Workers who take breaks of weeks or longer describe still having project access on their return (5 accounts across 4 threads, 2026).

Appen (CrowdGen). We have not built an evidence corpus for Appen, so this page makes no claim about how steady its work is. What CrowdGen's own pages say is that work is project-based and matched to your profile, and that hours are entered on the platform to get paid, with a dispute process for hours (crowdgen.com/payments, read 21 October 2026). Treat anything more specific you read elsewhere about Appen's work volume as unverified.

What workers actually compare Appen with

On Reddit the comparison people actually make is DataAnnotation against Outlier, and the Appen question is usually "sites like Appen": where else a worker can find the same kind of remote AI evaluation, AI training and data annotation work. Our evidence-based reviews answer that platform by platform, each built from dated Reddit comments that meet an evidence bar plus the company's own statements:

• Is DataAnnotation legit? (7,514 comments)
• Is Outlier legit? (Scale AI's contributor brand)
• Is Mercor legit?
• Is Remotasks legit?
• Side by side: Outlier vs DataAnnotation, Remotasks vs DataAnnotation, Mercor vs Outlier

If you are weighing Appen specifically, read its own pages (appen.com, crowdgen.com) for eligibility and payment terms, and treat third-party "Appen pays $X an hour" claims as unsourced until you see the project listing yourself.

If you are weighing Appen against Scale AI's contributor brand Outlier, see Scale AI vs Appen.

What about related platforms: Outlier, Mercor and Remotasks?

Outlier (operated by Scale AI) is the platform DataAnnotation workers most often mention as the alternative for language-model evaluation work. Our Outlier review covers pay, access and the account risks contributors report.

Remotasks is also a Scale AI brand, historically weighted toward image and structured-data annotation. See Is Remotasks legit? and Remotasks vs DataAnnotation.

Mercor recruits mostly domain experts (physicians, attorneys, engineers, finance specialists) for project contracts and publishes hourly rates on its listings. Our Mercor review covers what contributors report about continuity.

Experienced contributors keep accounts on more than one platform so that a dry spell on one is an inconvenience rather than an emergency. That is the single most repeated piece of advice in the threads our reviews are built from.

Which option is best for you?

If you are outside DataAnnotation's six countries: Appen's CrowdGen is the one of the two you can join. Apply to a specific project opening on crowdgen.com and expect a per-project qualification.

If you are inside those six countries and want to pick your own work: DataAnnotation's dashboard model lets you choose projects and get paid about a week after reporting hours, with the supply risk the review documents.

If monthly pay is a problem for you: CrowdGen pays by the 14th of the month for the previous month's work, so your first payment can be six weeks or more after your first task. DataAnnotation's roughly seven-day window is the shorter cycle.

If you want both: eligibility permitting, nothing stops you holding accounts on both, and on Outlier or Mercor as well. Many contributors do exactly that.

How to apply and get started

DataAnnotation. Apply on dataannotation.tech, take the starter assessment, then wait. Reported waits run from under 24 hours to weeks or months, some applicants never hear back, and there is no reported rejection message, so an unchanged dashboard is the only signal (our review, 10 accounts across 4 threads, 2025 to 2026; 5 accounts across 2 threads, 2023 to 2026). Our guides cover how to pass the DataAnnotation assessment, what the assessment is like, how long the review takes and the acceptance rate.

Appen (CrowdGen). Apply for one of the project openings on crowdgen.com; an account is created for you, you complete your profile, and most projects then have their own qualification process. Set up a payment method on CrowdGen before your first month ends, because payments are credited by the 14th for the previous month and existing Appen Connect contributors must re-add a method on the new platform (crowdgen.com/payments, read 21 October 2026).

Preparing for either. Both platforms' qualifications test the same five skills: writing or judging prompts for evaluation tasks, describing what an ideal response contains before judging candidates, reading rubrics precisely, grading consistently across cases that do not fit neatly, and writing an objective justification someone else could check. The AI Evaluator Certification teaches those five across 24 modules (30+ hours), with Kappa, the AI tutor, critiquing your reasoning in every lesson and never writing the work for you; the curriculum lists every module and the first one is free. No certificate replaces a platform's own assessment, and no platform requires one. If you are new to the field, start with how to become an AI trainer with no experience.