Skip to content
Labeling Jobs

Is AI training work legitimate, and how to spot the fakes

Overview · 8 September 2026

The category is real: Scale AI, Labelbox and Handshake all run one, and micro1 publishes peer-reviewed research on its own interviewer. That is exactly why the imitations work. Here is what a genuine platform does, the things that feel wrong but are normal, and the signals that mean stop, drawn from what the five platforms on this board actually publish.

The short answer

The work is real. Outlier is operated by Scale AI. Alignerr is Labelbox's expert network. micro1 has published a research paper on arXiv describing the AI interviewer it screens candidates with. Mercor publishes contractor documentation covering payments, holding periods and supported countries. These are companies with public policies you can read before you apply.

The category being real is precisely what makes the fakes work. "AI training job" now carries enough credibility to be worth imitating, and the imitations are aimed at people who have just learned the phrase.

The useful question is how to tell a real listing from one wearing its clothes.

What a real platform looks like

Every one of the five platforms on this board, without exception:

  • Publishes a rate before you commit. Alignerr publishes a rate on every listing before you apply. Outlier states you see the tasking rate before starting a project. micro1 and Mercor publish rates on individual listings. The rate may be a band, and it may be per task rather than per hour, but a number exists in writing.
  • Has a public company behind it with a website, a help centre or contractor documentation, and a name you can search independently.
  • Asks you to prove who you are, not to pay. Identity verification is universal here. Payment in the other direction is universal too: from them to you.
  • Names a payment rail. Stripe, Wise, PayPal, Airtm, ACH, or a platform wallet. Published, on a page, before you sign up.
  • States what it does not know. micro1 states that certification does not guarantee immediate placement. Handshake AI states there is no fixed timeline for a first project.

If a listing cannot produce those five things, do not trust it.

Things that feel wrong but are normal

Worth separating, because they cause people to walk away from real work and to relax around fake work.

An AI conducts your interview. Unnerving the first time, and completely normal here. micro1 built Zara and publishes research on it; Alignerr's process page states some of its jobs require a Zara interview too. See the Zara AI interview.

The assessment is unpaid. An unpaid AI interview before you know your rate is standard on micro1, and Mercor's own AI interview runs about 20 minutes by its account. Unwelcome, but not a scam signal.

They want your government ID. Standard across the field, and Outlier additionally requires a mobile phone number from your country of residence. Identity checks are how a platform meets its own compliance obligations.

You pass and then hear nothing. That is how matching works on these platforms: micro1 says certification is not placement, and Mercor matches you rather than letting you apply.

Tax paperwork before you earn. Mercor requires EU contractors to complete a W-8BEN during onboarding. Normal contractor administration.

The pay sounds too high. On this board rates run to $400 an hour for BigLaw-experienced attorneys and $245 to $280 for management consultants. High rates for scarce professional judgement are real. High rates for no qualifications are not.

The signals that mean stop

You are asked for money. Any amount, for any reason: a certification, a training course, a starter kit, equipment, a background check, a placement fee. This is the single most reliable signal. Every platform here pays you; none charges you. See do I have to pay anything to get AI training work.

You are asked to buy equipment and be reimbursed. A variant of the same thing, with a cheque that clears and then bounces after you have wired the difference back.

You are paid to click, then asked to pay to keep clicking. The US Federal Trade Commission calls these task scams, or gamified job scams. They usually start with an unexpected text or WhatsApp message offering "optimization" or "product boosting" work: rating products, liking videos, working through tasks in sets inside an app that shows a commission growing with every click. A small real payment may arrive early (the FTC says usually $5 to $20) to win your trust. Then, to unlock the next set or withdraw your balance, you are told to deposit your own money, usually in cryptocurrency. The balance was never real, and the deposit does not come back. Whatever the message calls it, this is not data labeling, and no platform here makes you fund a balance before it pays you. The FTC counted about 20,000 task scam reports in the first half of 2024, up from about 5,000 in all of 2023.

Banking credentials, not banking details. A real platform needs an account to pay into, which means an account number or a PayPal address or a Wise account. It never needs your online banking login, your card PIN, or a one-time passcode read out loud.

The whole process lives in a chat app. Recruitment that begins and ends in Telegram or WhatsApp, with no company domain, no application portal and no contract, is not how any of the five platforms here operate.

An offer with no application. Being approached with a firm offer for work you never applied to, especially with a rate attached, inverts how every one of these platforms works.

The domain is nearly right. Check it character by character against the company's own site. A suffix that merely ends with a known brand name is not the same as being that brand.

No rate anywhere. Not on the listing, not in the process, not on request. Every platform here publishes one somewhere before you commit.

Urgency. A same-day deadline on a decision that involves your identity documents or your bank details is a pressure tactic, not a hiring process.

Payment in cryptocurrency or gift cards. None of the rails used by the platforms here are of that kind.

Nothing exists outside the message. No help centre, no contractor documentation, no policies, no third-party trace of the company as an employer.

Where the real risk actually is

Focusing only on obvious fraud misses what more people actually run into.

On a legitimate platform, the realistic downside is investing unpaid time and receiving less than you expected: an unpaid assessment followed by no placement, a per-task rate that pays well below the advertised hourly band, a weekly submission minimum you were not told about, or a project that ends without warning.

There is also a documented dispute risk. Workers posting on Glassdoor and Reddit report Alignerr accounts deactivated with completed work still unpaid, and support going unanswered; Alignerr's own materials do not state what happens to unpaid completed work after a deactivation. We present that as a worker report rather than a platform statement, but it is the kind of thing worth asking about before you accumulate a balance.

How you actually get paid covers all of that, and the questions to ask before accepting.

What we could not establish

  • We have no data on how common fraudulent AI training listings are. The patterns above are reasoning from how the legitimate platforms operate, not measurements.
  • We cannot vet platforms beyond what they publish. Our company pages are written from each platform's own materials, and where a platform states nothing we leave the field empty rather than guess.
  • Being listed on this site means we found a real company with published policies and a working application link. It is not an endorsement, and it is not a guarantee of your experience.

Sources

Last updated 8 September 2026; interview length wording revised 25 September 2026; task scam signal added 2 October 2026. We are not affiliated with any of these platforms, we do not process applications, and we never ask you for money.

Labeling Jobs does not give financial or tax advice.

Questions

Is AI training work a real job?
Yes, as a category. Outlier is operated by Scale AI, Alignerr is Labelbox's expert network, and micro1 has published a research paper on arXiv describing the AI interviewer it screens with. These are companies with public policies, published rates and named payment rails. That the field is real is exactly why imitations of it exist, so the useful skill is telling a genuine listing from one wearing its clothes.
How do I know if an AI training job is a scam?
The most reliable single signal is being asked for money: a certification, a course, equipment, a background check or a placement fee. Every legitimate platform pays you and none charges you. Other stop signals are requests for banking credentials rather than banking details, a process that lives entirely in a chat app, a firm offer for work you never applied to, a near-miss domain name, no rate published anywhere, and pressure to decide the same day.
Is it normal to be interviewed by an AI?
On these platforms, yes. micro1 built its own AI interviewer, Zara, and publishes research on it, and Alignerr's process page states some of its jobs require a Zara interview as well. It is unpaid, and micro1's practice versions are labeled 18 minutes. Unnerving the first time, but not a sign that anything is wrong.
Should I give an AI training platform my ID?
Identity verification is standard across every platform on this board, and Outlier additionally requires a mobile phone number from your country of residence. Providing ID to a company you have independently verified is normal. Providing it to a recruiter operating only through a chat app, with no company domain and no application portal, is not.
What is the realistic downside of legitimate AI training work?
Not fraud, usually, but unpaid time and disappointed expectations: an unpaid assessment followed by no placement, a per-task rate that pays well under the advertised hourly band, a weekly submission minimum you were never told, or a project that ends without warning. There is also dispute risk; workers report Alignerr accounts deactivated with completed work unpaid, and Alignerr does not publish what happens to that work.

Platforms covered here

Put this into practice

Every listing shows its pay and who it is open to.

Browse all jobsGet new jobs on TelegramGet new jobs on Discord