Working training AI: I compare the 6 biggest platforms (tier list)

Training artificial intelligence pays between 20 and 60 dollars an hour. It is not one of those internet promises: it is what thousands of people are earning right now from home, evaluating AI model responses and teaching them to be more accurate. The problem is choosing a platform, because there are dozens, they all say the same thing on their homepage, and some keep you waiting months with no projects while others pay you every week. So here is my personal tier list of six of the big ones: the good, the bad, and what nobody tells you before signing up.

Most people rank these platforms by what they pay per hour. And that is a mistake, because a spectacular rate on a platform where there are never any projects available is worth exactly zero. What really determines what you end up earning is the multiplication of three things: how much they pay per hour, how many hours you can actually work per month, and how often they pay you. A platform that pays 40 dollars an hour but only gives you work two weeks a year is below one that pays 25 and gives you work every month. I built the list with that logic, and that is why some of the famous ones will surprise you at the very bottom.

We start at the bottom, and I want to be fair: Toloka is not a scam, far from it. It is a real platform that has been running for years and pays what it says. The problem is that it fails on all three legs of my formula at once: the rates are among the lowest in the industry, task availability is erratic and depends heavily on your country and language, and the tasks available tend to be mechanical (classifying and labeling), which is exactly the worst-paid work in this whole industry. When would I use it? Only if you have never done any of this and want to get your hands on an annotation task to understand what it is about. As a school it is fine; as a source of income, there are five better options in this same list.

First surprise: Outlier, the best known of them all, the one in every ad and video, lands in tier C for me. And it is not a whim. Outlier has the highest rate ceiling on this list, no argument: there are specialists earning figures many full-time jobs would envy. And it has an enormous variety of projects and languages, which makes it accessible from almost any country. But when you apply my formula, it falls apart. Its big problem is volatility: they have moved to an on-the-fly project assignment system, and the result is that you can have a week of work and then three weeks of absolutely nothing, even with good ratings. Add slow support (people reporting waits of days or weeks to resolve payment issues) and account closures with no explanation and no way to appeal. Its reputation among workers has been declining for a while. You can make money on Outlier, and a fair amount. But it is the definition of feast or famine, and if you want stable income, it will drive you crazy. The most famous is not the best.

Clickworker rises to tier B for a very specific reason: it is the only one on the list where getting in is practically guaranteed. It accepts almost any country, has a mobile app, has tasks of all levels (even in-person mystery-shopper types), and does not put you through brutal elimination exams to let you start. It also has a very interesting secondary door: from inside you can unlock access to UHRS, an external platform with thousands of microtasks, by passing a small test. That multiplies your work volume at once. I explain it in depth in the Clickworker guide and how to unlock UHRS. Its limit is obvious: the rates are the most modest of all those ahead of it. You will not see 40 dollars an hour here, or anywhere close. So why tier B and not C? Because of the second leg of my formula: here there is ALWAYS something to do, no three-week droughts staring at an empty screen. It pays less, but it pays consistently, and that is worth far more than people think. It is the best first platform for a beginner, and a good cushion for the slow days of the others.

Serious stuff now: tier A for DataAnnotation. Rates that start around 20 dollars an hour and rise notably on technical or programming projects, with a work queue considerably more stable than Outlier: fewer silent droughts and more predictability. And if the rate and stability are so good, why is it not higher? Because of the entry door, which is the toughest on the whole list. To get in you have to do unpaid sample tasks, and the selection process is an absolute black box: they do not tell you why they accept you, they do not tell you why they reject you, and most people end up waiting months with no answer. It is the classic "if you get in, congratulations; if not, you will never know what went wrong" platform. My advice: apply anyway and take the entry test very seriously, because it is literally the best-paid hour of your life if you pass it.

It shares tier A with Alignerr, backed by Labelbox, an established data-infrastructure company that works behind much of this industry. It works in the way that, for me, is the fairest of all. Instead of a generic queue of tasks dropping on you, Alignerr has a project board, similar to a job board: you go in, see the open projects with their hourly rate written BEFORE you apply, pick the one you want, and sign up. Community-reported rates range from about 15 to 60 dollars an hour depending on the project, and the entry barrier is considerably friendlier than DataAnnotation. Its weak point, and why it stays in tier A and does not go higher: since you work by project, when one ends you have to go back to the board and apply again, and there can be gaps between one and the next. Even so, for rate transparency and weekly pay, if I had to recommend ONE to start today without being an ultra-specialized profile, it would be Alignerr.

We reach tier S, which in my list has a single name, and it is not the most famous by a long way. It is here because it is the only one that scores high on all three legs of my formula at once: high rate, real access to work, and a model that does not treat you as disposable labour. Number one is micro1. micro1 is not a microtask site where you peck at odd little jobs. It is what they call a data lab: they work directly with cutting-edge AI labs, building realistic training environments and evaluations for their models. And for that they need genuine human experts. How do you get in? On their site they have an expert opportunities section, where you apply to open projects according to your specialty. You pass their verification process, which includes an interview with their AI system, and if selected you work on projects for those labs. Why is it my number one? Because they pay you for your knowledge, and there the rates have nothing to do with labeling images. They also work with global talent, so it is one of the most direct routes for someone in Spain or Latin America to access international lab projects that pay in dollars.

A recap of the list: in tier S, micro1. Tier A, DataAnnotation and Alignerr. Tier B, Clickworker. Tier C, Outlier. And tier D, Toloka. The strategy that works in this sector is having several running at once, because the inconsistency is not the fault of any one platform: it is the nature of the whole industry. When one goes into a drought (and it will), the others carry your month. My recommended combination if you start today: Alignerr as the base, because you get in relatively fast; micro1 if you have a specialty to offer, because that is where your knowledge is worth most; and Clickworker as a cushion for the days the others are empty. Keep DataAnnotation as a medium-term bet while you get through its entry test.

The usual honest reflection: this is among the best-paid flexible remote work that exists today, it is completely real, and there are people making very serious money. But it is not a fixed job and does not pretend to be: it is project work, with its ups and downs. Anyone selling it to you as a guaranteed salary is deceiving you. Even so, with expectations where they should be and three platforms running, it is one of the best opportunities out there right now. You have access to these platforms in our AI training category, and the rest of the verified methods in the complete directory, each with its own analysis.