What really decides how much you get paid to train AI
There is something nobody tells you about AI training work: what really decides how much you get paid is not what you know about AI. It is what you know about your own field. Programming, medicine, law, finance, a technical trade, a specific language.
There is a very logical reason for this, which I explain in the next section, but hold onto the idea: if you have a specialty of any kind, even if it seems to have nothing to do with this, you are probably worth a lot more than you think.
In this guide I cover how Meridial works, some tips and tricks, and what it really pays.
What Meridial is and why it pays what it pays
Meridial did not come out of nowhere. Behind it is Invisible Technologies, a company that has spent years working with AI labs, and Meridial is the marketplace where they distribute the projects those labs commission out to outside people.
The work is the usual kind in this sector: writing example questions and answers, evaluating and ranking what the model responds, correcting its reasoning when it gets things wrong, and deliberately poking at it to find where it fails.
And here is the key thing to understand: an AI model today writes well about almost everything. The problem is it also says nonsense with the exact same confidence it says the truth. To fix that, they need someone who actually knows that specific subject and can catch it: someone who knows medicine and spots that what it just said about a drug is wrong, someone who knows law and sees that a clause is not drafted that way, or someone who codes and notices that piece of code looks right but does not compile.
That is why what puts you at the top of the pay range is your field, whatever it is. And that is why the single most profitable piece of advice in this guide is: do not present yourself as someone interested in AI, because everyone says that and it lumps you in with the generalists, which is exactly where the low rates live. Present yourself by what you actually know how to do.
How to get in: the 6 gates (and the one-shot one)
The application process is long, and it is worth knowing that before you start. There are basically six gates: the application with your résumé, an initial aptitude test, an English assessment, the specific test for the project you apply to, identity and address verification, and finally signing the work agreement. Some of these tests are proctored, with your camera and microphone on and your screen shared. Do not let it catch you off guard.
The English assessment: you only get one shot
This is not just a formality: the level you score there determines which projects you get accepted into and which ones you do not. Some people who scored a lower-intermediate level ended up not qualifying for a single one, and worse still, you cannot retake it.
So listen carefully to this: that test is not something you do on a Tuesday night while eating dinner. It is the gate that decides everything that comes after, and you only get one shot. Do it rested, with time to spare, somewhere quiet.
And if your English is just so-so right now, the honest recommendation is: do not take it yet. Spend a few months improving it and apply afterward, because going in at a low level locks you in at that level with no way to change it.
Where everyone gets stuck: identity and address
The other gate where a huge number of people fall is the fifth one, verifying identity and address. These are the three most common mistakes.
First, the biggest one: the address you enter on the form has to match exactly what appears on your bank statement. Exactly, with the same abbreviations and the same format, and it matters because it cannot be edited afterward. Open your statement, copy it as-is, and double-check it before submitting.
Second: your proof of address has to be a bank statement or a utility bill in your name. An ID document does not count as proof of address, and plenty of people have sent one and gotten rejected.
Third: photograph your documents properly: flat, in daylight, with all four corners visible. It sounds trivial and it is one of the most common rejection reasons there is.
The projects available and what they pay
There are hundreds of projects open at once, spread across about a dozen areas: languages and linguistics, sciences, business, engineering, programming, medicine, finance and math, and creative work.
Notice something that ties back to what we said at the start: the biggest area by far is languages. Out of every five projects, almost two are languages and linguistics. That means speaking a language other than English well is a specialty that pays off handsomely here.
The good thing about this platform is that you do not have to guess anything: every project lists its rate and the country or region you can work from.
Two tips about the published rates
First: the rate is a range, not a promise. A high figure listed means someone on that project earns that much, not that you will be paid that. Your specific rate gets confirmed in the invitation once you have been matched, before you sign anything, and you can turn it down if you want.
Second: always check what requirements a project asks for before getting into the process. You can spend an entire afternoon preparing an application for a project that does not accept people from where you live, and not even the best résumé in the world fixes that.
The 3 things that decide your rate
What decides whether someone gets paid the bottom of the range or the top. Three things, from least to most uncomfortable.
1. Your experience
The expected one. More years, more demonstrable experience, higher rate.
2. Your specialty
This is the one that really moves the needle. However unusual your field seems, it counts here, because the more specific it is, the fewer people know it. And scarcity is what drives the price up.
3. Your location
This is the uncomfortable one, which is exactly why it is stated plainly. The platform itself says the rate depends on location, and there are accounts of people being offered very different amounts for the same project depending on where they worked from. That is just how it is, no sugarcoating it.
So do not let it catch you off guard when your offer arrives. And specialize even further in your field, because of the three levers, specialty is the only one you can actually move yourself.
The honest part: why it has such a bad rating
If you search for Meridial reviews online, you will find a terrible rating, one of the worst among platforms of this kind. And before you decide it is a scam and close the tab, those reviews need a careful read.
The vast majority of those complaints come from people who never actually got to work. They are complaints about the application process: that it takes forever, that they got stuck in verification, that they have been waiting weeks with no word, that they opened a ticket and it got closed automatically with no human ever responding.
All of those complaints are real and legitimate, to be clear. The application process on this platform is brutal, which is why we gave it an entire section. But those are complaints from someone trying to get in, not from someone who did not get paid. And those are two very different things that people constantly mix up.
The honest read is this: the risk here is not that you work and do not get paid. The risk is that you sink a lot of time into the application process and never actually get in. And you need to know that before you start, because nobody pays you for that time.
Three more things almost nobody mentions
Getting into the network does not guarantee you work. The platform says this itself. Your profile just sits there and you get notified if you match something, and that match can take weeks or never arrive.
Projects run out. This is not a job, it is a dated assignment, and once it ends you go back to looking for another one.
And the good one, which almost nobody mentions: once you are inside a project, the time you spend reading instructions, watching recorded training, and in onboarding meetings counts as paid time. That is not at all common in this space; normally all of that part comes out of your own unpaid time.
Who it is for, and who it is not
So who is this for? I will split it fairly, because that is the honest way to do it.
If you are starting completely from scratch, with no specialty and so-so English, this is not your platform. I am not saying that to discourage you, I am saying it so you do not waste three weeks on an application process you will probably not pass. For that starting point there are much more accessible platforms, covered in the best platforms to train AI comparison.
But if you have a genuine specialty, whatever it is, and you can hold your own in English, then this is one of the best options out there right now. You get paid for knowing what you already know, with a signed document in front of you stating the rate, and payments every two weeks.
The final piece of advice, the one that actually changes the outcome: make your specialty the first thing anyone reads on your profile, not your interest in artificial intelligence, which honestly matters very little here.
If you want to compare with other platforms in the same space, you have micro1 and DataAnnotation covered in depth, and the rest of the no-client-hunting platforms in 10 platforms to make money without finding clients.
