AI training work FAQ: 12 questions everyone asks

Yes. The platforms that list AI training roles are real companies paying real people to do real work.

micro1 is venture-backed ($35M Series A at a $500M valuation, September 2025). It pays in USD twice a month through Deel.

Mercor pays weekly via Stripe (or Wise where Stripe is unavailable), processed every Wednesday. It has over 5 million signups, though far fewer are actively working.

Terac was founded in 2025 in San Francisco, backed by $9M in seed funding from Emergence, SignalFire, Audacious, SV Angel and Z Fellows. It holds a 5.0 rating on both Trustpilot and Google, though the review sample is still small for a young platform.

AfterQuery operates a smaller catalogue with a narrower skill focus.

All four classify workers as independent contractors, not employees. That means no benefits, no guaranteed hours, and you handle your own taxes. But the work itself is not a scam — you do tasks, you get paid.

The gap between signing up and getting paid work catches people off guard. Passing a screening assessment does not guarantee a steady flow of tasks. Some people apply, pass, and then wait.

Across the 597 open roles this site tracks, listed hourly rates span $2–$400 an hour. Most roles fall between $15 and $75/hr.

The range reflects the work. A bulk data labelling task with thousands of openings pays $6/hr. A niche legal or medical review that needs specific credentials pays $200+/hr.

What matters more than the listed rate is how many hours of work are available. A $75/hr role with two hours of available work a week pays less than a $25/hr role with forty. Listed rates are real, but they are not income until you are actually doing tasks.

Not every role pays by the hour. Of the 597 open roles, 540 list an hourly rate, 40 pay a fixed amount per session, and 17 pay per task.

For the full breakdown by platform, see what AI training work pays.

1. Browse open roles and find ones that match your skills
2. Apply on the platform — each has its own screening process
3. Pass the assessment (a coding test, writing sample, or domain quiz depending on the role)
4. Wait to be matched with available work

Every platform screens applicants. On micro1, that is a skills assessment tied to the specific role. On Mercor, it often includes a technical assessment and may add an interview. On Terac, some roles begin with the evaluation itself — the screening is the task, and you are paid for completing it.

No platform charges applicants. If something asks for money, it is not one of these platforms.

For a longer walkthrough, see how to get started with AI training work.

No degree is strictly required for most roles. What matters is whether you can do the specific work the role asks for.

That said, credentials affect which roles you can access and what they pay:

• Generalist roles (data labelling, basic evaluation) need clear thinking, attention to detail, and the ability to follow annotation guidelines. No degree required.
• Technical roles (code review, software engineering tasks) need demonstrated coding ability. A CS degree helps but is not required if you can pass the assessment.
• Domain expert roles (legal review, medical evaluation, physics) typically need the credentials the posting describes. A listing that says “physics PhD required” means it.

The screening tests are the real gatekeeper, not your resume. Strong written English and the ability to follow detailed instructions consistently are the baseline for most platforms.

Yes, but availability varies by platform and by role.

Of the roles this site tracks, the majority are open globally. Some are restricted to specific countries — the role listing says so when they are. You can filter roles to see what is available in your location.

micro1 pays in USD through Deel to contractors in dozens of countries. Mercor pays via Stripe or Wise depending on your country. Terac requires government ID verification during onboarding and pays via PayPal globally.

The main variable is not whether a platform operates in your country, but whether the specific role you are applying for accepts applicants from there. Country restrictions are set per role, not per platform.

AI training platforms classify you as an independent contractor. That means:

• No taxes are withheld from your pay — you receive the full amount
• You are responsible for reporting and paying your own taxes
• In the US, if you earn $600 or more from a single platform in a tax year, you should receive a 1099-NEC form (this threshold rises to $2,000 for the 2026 tax year)
• US self-employment tax is 15.3% (Social Security + Medicare) on top of your income tax rate
• The Qualified Business Income deduction lets eligible self-employed filers deduct 20% of qualified business income

If you are outside the US, your country’s tax rules for freelance or contract income apply. Most countries tax foreign-sourced freelance income. Check your country’s rules for self-employment or freelance work — the specifics vary widely.

Keep records of what you earn and when. Set aside a portion of each payment for taxes rather than spending it all.

Task availability fluctuates across every platform. A role listing thousands of openings does not mean thousands of hours of work are waiting for you specifically.

Common reasons for an empty queue:

• The project finished or paused. AI training work is project-based. When a client’s project wraps up, the tasks disappear.
• Your quality scores. Most platforms route more work to higher-rated workers. If your accuracy drops, your queue thins.
• Time of day and timezone. Some tasks are released in batches at specific times.
• Oversupply of workers. When a platform onboards more people than it has tasks for, everyone’s share shrinks.

The standard advice from experienced workers is to maintain active accounts on two or three platforms rather than depending on one. Task droughts happen to everyone, and diversifying is the main hedge.

Yes. These platforms classify you as an independent contractor, and none of the four this site tracks has an exclusivity clause that prevents you from working elsewhere.

The NDAs you sign are about protecting the AI client’s data — what models you are training, what prompts you see — not about preventing you from doing similar work on another platform. You cannot discuss the specific projects, but you can work on as many platforms as you can manage.

The practical limit is quality, not policy. Spreading yourself across too many platforms can degrade your work on each one, which lowers your quality scores, which means fewer tasks. Two or three active platforms is what most experienced workers settle on.

Browse the micro1 roles, Mercor roles, Terac roles, and AfterQuery roles to see what each has open.

It depends on what you do and what you are looking for.

If you are unsure, apply to more than one. The applications are separate, the screening is free, and nothing stops you from working on multiple platforms.

For a detailed side-by-side comparison of the two largest platforms, see micro1 vs Mercor.

For most roles:

• A laptop or desktop with a reliable internet connection
• A browser (most platforms run entirely in the browser)
• Fluency in the language the role specifies

Some roles have additional requirements — a specific OS, a minimum screen size, a webcam for proctored assessments — but these are stated in the listing. The default is a computer and a browser.

A phone is not enough. The annotation interfaces, coding environments, and evaluation tools these platforms use are built for desktop browsers.

Not yet, and probably not soon, but the work will change.

AI training work exists because current models need human judgement to improve. Humans decide whether an AI answer is correct, whether it followed the prompt, whether it chose the better response. That judgement loop is what these roles are.

As models get better, some of the simpler tasks — basic data labelling, obvious-answer evaluation — will need fewer humans. But the harder tasks — nuanced domain review, edge-case evaluation, red-teaming — grow in demand as models become more capable, because the remaining errors are harder to catch.

More likely than elimination is a shift: fewer people doing basic annotation, more people doing expert-level evaluation, and pay diverging further between the two.

Payment methods and schedules vary by platform:

• micro1 pays in USD, twice a month. The hiring company pays you through Deel.
• Mercor pays weekly via Stripe, processed every Wednesday around noon PST for the prior work week. Stripe enforces a 7-day hold on your first payout. In countries where Stripe is unavailable, Mercor uses Wise instead.
• Terac lets you cash out on demand once your balance reaches $10. Methods include PayPal, Venmo, Visa prepaid card and direct cashout, with PayPal available in every country.
• AfterQuery payment details are stated in the role listing.

All four platforms pay in USD. If you are outside the US, your payment provider handles the currency conversion at payout time.

No platform charges you to receive payment. If something asks for a fee to “process” your earnings, it is not one of these platforms.