Outlier, DataAnnotation, micro1 or Mercor: where to apply first

Outlier, DataAnnotation, micro1 and Mercor are four of the best-known names in paid AI training work, but they run two different businesses. Outlier and DataAnnotation are marketplaces: you qualify, then take work when there is work. micro1 and Mercor are matchmakers: you apply for a specific contract and work it until the project ends. Which to try first depends on what you bring, where you live and how much unpredictability you can carry.

Marketplaces and matchmakers#

On a task marketplace, the platform keeps a pool of qualified contributors and releases work project by project. You never meet the client; you work as much as the queue allows and are paid for the time or tasks you complete. The advantage is speed and flexibility. The cost is volume you do not control: when a project ends or a customer leaves, the queue can empty overnight, often without explanation.

On a contract matchmaker, you apply to a listing for a specific project and are screened, typically by an AI interviewer. If you are selected, you join with set expectations on hours and quality. The advantage is predictability while the contract runs and, for credentialed people, higher-value work. The cost comes at both ends: a match can take a long time to arrive, and when the project ends you are back to applying.

Neither model is safer. They fail in different ways, so the strongest position is to hold one of each.

The four platforms, one at a time#

Outlier: broad entry, uneven volume#

Outlier is the contributor platform of Scale AI, one of the largest suppliers of training data to AI labs. It recruits for generalist writing, coding, mathematics, the sciences and many languages.

You build a profile with up to 10 skills, pass a short skills screening and verify your identity with a government ID. Then you wait to be matched to a project and pass its onboarding before paid tasks appear. Reviewers score your work against each project's guidelines, and weak scores can end your time on a project.

Pay is set per project, usually by the hour, and depends on the work, your expertise and often your location; current ranges are on the Outlier profile. Payouts run weekly through PayPal or Airtm, with ACH bank transfer where available.

The trade-off is volume. Its queues follow its customers' projects, and the empty queue is the platform's best-known complaint: when a project ends, the dashboard can sit idle for days or weeks.

DataAnnotation: one assessment, then your own hours#

DataAnnotation keeps entry simple. The application is an online Starter Assessment that tests how carefully you follow instructions, reason and write, with separate versions for specialist work such as coding. You get one attempt.

If you are accepted, paid projects appear on your dashboard and you choose what to work on and when, with no shifts; passing further qualifications opens more projects. Advertised starting rates are higher for coding and specialist work, and current figures are on the DataAnnotation profile. Payouts go through PayPal.

The weak point is opacity. Rejections usually arrive as silence, so you will not learn why. Keep applying elsewhere while you wait; our guide to passing AI training assessments covers how to prepare.

micro1: an AI interview first#

micro1 began as a platform for hiring software engineers vetted by AI and now also supplies human experts to AI labs. Its signature is Zara, an AI recruiter that runs a recorded, spoken interview built from the listing's skills and your CV. Technical roles can add a coding challenge.

You apply to specific roles, not to the platform. Passing the interview and an ID check puts you in a talent pool, not on a project. If you are matched, you join a defined project and are paid by the hour or per approved task; the micro1 profile tracks current rates and payouts. It suits engineers and domain specialists who interview well and prefer a defined engagement to an open queue.

Mercor: contracts built on credentials#

Mercor works the same way, with a heavier tilt toward credentialed professionals. You upload a résumé, apply to listings and complete a short AI video interview; some roles add a domain assessment. Listings run from software engineering and generalist writing to law, medicine, finance and consulting, and the best-paid ones ask for advanced degrees, licenses or years in the field.

Hourly offers cap your weekly hours, and a tracker that takes periodic screenshots logs your time. Current rates and payout methods are on the Mercor profile, and our guide to AI interviews at micro1 and Mercor covers the format.

Side by side#

Applying from Africa: what we know and what we don't#

Country rules are among the least transparent parts of this work, and they change, often quietly. Three things decide whether you can work on a platform from Lagos, Nairobi, Accra, Johannesburg or Cairo.

The first is the platform's list of supported countries. The second is the payout rail: a platform can pay you only if its payment partner can reach your account, and what PayPal lets you receive varies by country. The third is the project, since some contracts require a country of residence, a professional license or a language.

For these four, as of October 2026:

• Outlier works from a list of permitted countries. Egypt and South Africa are on it; Nigeria, Kenya and Ghana are reportedly missing from the version updated in July 2026. Where you are eligible, Airtm gives a second payout route where PayPal is impractical.
• DataAnnotation reportedly accepts applications from six countries only: the United States, Canada, the United Kingdom, Ireland, Australia and New Zealand. As far as public evidence shows, it is closed to all five countries here.
• micro1 does not list Nigeria, Kenya, Ghana, South Africa or Egypt among the countries it cannot hire from, and pays through Deel. Mercor lists Nigeria, Ghana, South Africa and Egypt as supported and pays through Stripe; Kenya appeared on its unsupported list, though one Mercor page has suggested otherwise. Both decide listing by listing, and many Mercor listings require residence in the US, UK or Canada, which matters most for professionals whose licenses are local.

Every platform profile carries the details and the date we last checked them.

Three rules hold everywhere. Apply from the country where you actually live, under your own identity, since platforms verify both. Use the same legal name on your platform account, your ID and your payout account, because a mismatch can hold up a payment. And confirm you can receive money before you work, not after. Our guide to getting paid from Africa covers the options.

What the Meta and Scale deal taught contributors#

The clearest lesson of the past two years did not start on a contributor platform. In June 2025, Meta invested in Scale AI, Outlier's parent, in a deal widely reported at $14.3 billion for a 49 percent stake. Scale's founder and chief executive, Alexandr Wang, left to join Meta.

The headline was the investment. What followed was customer flight. Within days, reports said Google planned to wind down its work with Scale, and OpenAI said it was phasing out a relationship it had already been scaling back. In July 2025, Scale cut about 14 percent of its staff, roughly 200 employees, and ended work with about 500 contractors. Its interim chief executive at the time said the company had ramped up its generative-AI capacity too quickly.

For Outlier contributors, a shift like that shows up in the queue. A marketplace's volume is borrowed from its customers: when a lab of Google's size pulls back, its projects wind down, and the people trained for them lose work through no fault of their own.

The lesson reaches beyond Scale. Demand for this work comes largely from a small number of AI labs, so a matchmaker's contract can end for the same reasons a marketplace's queue empties. In March 2026, Mercor was hit by a supply-chain attack through LiteLLM, an open-source software library. Meta then paused its work with the company, leaving contractors on Meta projects unable to log hours. Lawsuits allege that contractors' identity documents and interview recordings were exposed; Mercor denies wrongdoing.

Concentration is the risk; holding more than one platform is the hedge. For the wider picture, see the state of AI training work in 2026.

Where to apply first, by profile#

One rule covers most people: start where your strongest credential is scarce, and pair it with a platform that can give you work quickly.

Generalist writer. If you live in a country DataAnnotation accepts, start there, with Outlier as your second door. In most of Africa, look first at micro1's generalist listings and, where it is supported, Mercor's; language and data platforms such as OneForma and Appen open work project by project. Generalist English is the most crowded pool, so polish matters more than speed.

Software engineer. Start with micro1, whose screening grew out of vetting engineers, then Mercor for software and evaluation contracts. Where DataAnnotation accepts applicants, its coding track is the self-paced complement. Our guide to coding for AI labs covers what reviewers look for.

STEM graduate or PhD. Start with Outlier if it accepts your country; it recruits widely across mathematics and the sciences. Add Mercor if you hold a doctorate or deep research experience; its expert listings pay for depth.

Doctor, lawyer or finance professional. Start with Mercor, whose listings lean hardest toward credentialed professionals. Read the eligibility line closely, since a license often ties a contract to a country. micro1 is the second application.

Student. Start with DataAnnotation or Outlier, whichever is open to you: DataAnnotation's generalist track accepts an in-progress bachelor's degree, while Outlier's FAQ asks for at least an associate degree. Matchmakers mostly hire for experience, though some Mercor listings target students and recent graduates of top-ranked universities. If you have a technical streak, try ASV's own paid projects, from task authoring with rubrics to software tasks.

Sources#

• Frequently asked questions (Outlier)
• Flexible Working Guidelines (Outlier)
• Frequently asked questions (DataAnnotation)
• Leverage your expertise to train next-generation AI models (micro1)
• Supported countries for payment (Mercor)
• Meta invests in Scale AI while OpenAI and Google step back (Pure AI)
• Scale AI lays off 14% of staff amid revamp following Meta's $14.3bn investment (Outlook Business)
• After data breach, $10B-valued startup Mercor is having a month (TechCrunch)