Is Alignerr Legit? What 3,988 Reddit Comments and 60 Live Listings Show

The short version

• Alignerr is real. It is Labelbox's contributor-facing brand, and it does pay people.
• Some projects pay per approved task, others track your hours. Establishing which one you are on before you start work is the most important thing on this page.
• On per-task projects, approval gates payment. If a project pauses before your work is reviewed, it may never be reviewed, and unreviewed work is unpaid. 25 different accounts across 18 threads describe losing work exactly this way, and the reports are increasing rather than fading.
• Being removed from a project is the most reported problem in subreddits Alignerr does not moderate: 11 accounts across 7 threads, including one worker dropped mid-task.
• Published rates today: $20 to $50 per hour for generalist task authoring, $60 to $120 per hour for roles requiring an established profession such as software engineering, accountancy, OpenSees or OpenFOAM. Nine distinct roles are live as of 30 July 2026.
• Plenty of people are paid well, and that is equally documented. Both things are true at once.

The useful question is not whether it is a scam. It is whether you understand where the risk sits before you start. It sits with you, not with the company.

What Alignerr is

Alignerr is how Labelbox recruits and pays the people who do its AI training and evaluation work. Their own listings carry the line "Alignerr | Powered by Labelbox," and 65 different people across 27 Reddit threads make the same connection.

You sign up, complete screening for a given project, and then work tasks: comparing model responses, writing prompts, transcribing, annotating video, or authoring evaluation tasks in a technical field. It is the same category of work as AI evals generally.

How we checked

We read 62 Reddit threads containing 3,988 comments from 2,004 different accounts, dating from July 2024 to July 2026. Every claim below had to be stated by at least two different accounts in two different threads before we would print it, and every claim carries the period its evidence comes from.

We deliberately read in two passes. The first covered Alignerr's own subreddit. The second covered places Alignerr does not moderate, because a company's own forum is not neutral ground. That turned out to matter: one worker reports being banned from the Alignerr subreddit after complaining about withheld pay, which is exactly the kind of filtering that would make a single-source guide misleading. Every finding below appears in both passes.

Where people disagree, we print both sides. Where the evidence is mostly old, we say so instead of presenting it as current.

We also track Alignerr roles on our own AI evaluation job board, so the section on what is open right now is our own data rather than anyone's recollection.

What Alignerr actually pays, according to Alignerr

People quote wildly different numbers in forum threads, and none of them can be checked. So rather than repeat any of it, here is what the company itself publishes. We list Alignerr's roles on our job board, and every one carries a rate.

As of 30 July 2026 there are 60 live Alignerr listings, which are 9 distinct roles, posted across 14 country eligibilities:

Three things worth reading off that table.

The high rates are professional work, not annotation. Everything above $60 per hour asks for a specific profession: software engineering, accountancy, structural engineering with OpenSees, computational fluid dynamics with OpenFOAM. You are being paid to author evaluation tasks in a field you already work in. If you are picturing entry-level labelling, the relevant rows are the $20 to $50 ones.

Sixty listings is nine jobs. The same role is posted once per country. Eligibility across the set covers 14 countries including Germany, the United States, India, the United Kingdom, Australia, Canada, Brazil, Singapore, the Netherlands, the Philippines, Colombia and Ireland.

A published rate is not an earnings figure. It is what Alignerr says a role pays. What reaches your account depends on how many of your tasks get reviewed and approved, which is the subject of the next section.

We began tracking Alignerr on 25 July 2026, so this is a dated snapshot rather than a trend. We are not going to tell you their hiring is up or down on five days of history.

How the money actually works, and why it matters

Alignerr does not use a single payment model. It uses two, and which one applies to you depends on the project you are assigned.

Some projects are time-tracked. Workers describe logging hours and using a timer, and at least one project runs through Hubstaff, a time-tracking tool. One person said in December 2025 that they had stopped taking Alignerr work unless it was paid per task, precisely because they disliked the tracking. That single sentence tells you plainly that both arrangements exist.

Other projects pay per approved task. One worker described a project in April 2026 paying 260 to 300 per task rather than by the hour. Another argued in December 2025 that the complexity of a particular project made hourly payment impossible for it.

Nineteen different accounts across 17 threads say some version of "it depends on the project." So the honest instruction is not that Alignerr pays per task. It is: find out which model your project uses before you do any work, because the risk is completely different.

On a time-tracked project, logged hours are the record. On a per-approved-task project, approval gates payment, and that is where the losses happen:

• If a project is paused or closed while your work is in the review queue, the work may never be reviewed, and unreviewed work is unpaid.
• If a task is reviewed and failed, you are not paid for the hours it took.
• Preparation, which several people describe as hours of instructional video and reading, is not itself a paid task.

Twenty-five different people across 18 different threads describe losing work this way, and the reports are not fading with time. There were 8 in 2024, 9 in 2025, and 10 already in 2026 despite our sample containing fewer recent threads.

Here is the other half, and it is equally real. In the same threads, in the same months, people report being paid steadily and well: earnings in the thousands, a documented span of six weeks with consistent payment, ordinary project work with no problems at all. One person who has done both technical and non-technical work put it this way in July 2026: it is not a perfect place, but it is not a scam. The positive accounts survive outside Alignerr's own subreddit too, including someone in December 2025 who had been suspended by a competitor and said Alignerr treated them well.

Both things are true at once, and that is the honest finding. The arrangement works when a project runs to completion and your work gets reviewed. It fails you when a per-task project stops early, and in that case you carry the risk rather than the company. One worker put the comparison directly in April 2026: some competitors pay freelancers for completed work even when the client stops paying them, and their read was that Alignerr does not.

What else people consistently report

You can be removed from a project, sometimes mid-task. One worker was dropped after two weeks without completing a task. Another cleared the screening and was never added to the project at all. A third, writing in September 2025, had completed four full sets of tasks with a fifth around 85 percent done when they were removed. Removal is the single most reported theme outside Alignerr's own subreddit: 11 different accounts across 7 threads.

Screening can be repeated, and some of it looks like work. In July 2026 someone described passing a qualification after hours of preparation, then being presented with an updated qualification the next day whose questions did not link to the correct guidelines. Separately, two people in outside subs describe qualification assignments substantial enough that they questioned whether they were doing unpaid production work. We cannot confirm what happens to that output, only that the assignments are long and unpaid.

Work availability is uneven. This is the one place to be careful. Complaints about dry spells and missing projects are real, but they cluster heavily in 2024, with far fewer in 2025 and almost none in 2026. That pattern is consistent with the situation having improved, so we are not going to present a two-year-old complaint as today's condition.

Organising has worked at least once. In May 2026 a group of workers published an open letter about unpaid, unreviewed tasks on a paused project. The person who wrote it later edited the post to say the problem was solved: once people got organised, the tasks were reviewed and they were paid. That is one documented case rather than a reliable remedy, but if you are in that position it is the only thing in two years of threads that we can show actually working.

What we could not verify

What any individual actually earned. Forum figures vary by more than an order of magnitude and none of them can be checked, so we do not repeat them. The rates in this guide are Alignerr's own published rates, taken from live listings on the date shown.

Claims about intent. A lot of people believe tasks are failed deliberately to avoid paying. We can see the effect, and the effect is real. We cannot see the motive, so we describe the mechanism and leave the conclusion to you.

Third-party ratings quoted second-hand. A widely repeated review-site score appears in these threads. We have not checked it ourselves, so we are not repeating it.

If you decide to try it

• Find out whether your project is time-tracked or paid per approved task before you do any work. It changes everything else on this list.
• On a per-task project, treat unreviewed work as money at risk, not money earned, until it clears review.
• Expect preparation time to be unpaid, and factor it into what the rate really means for you.
• Keep working on a project once you are on it, since inactivity can cost you the seat.
• Do not rely on a single platform. The most common advice in these threads, from people on all sides of the argument, is to be on several. Our job board tracks 15 of them.
• If a project is paused with your work unreviewed, find the others in the same position. That is the one thing on record as having produced payment.

Help us keep this accurate

These platforms change constantly, and a guide that is right today can be wrong in three months. If you have worked with Alignerr and something here does not match your experience, tell us with roughly when it happened, and we will update the guide and note what changed.

Method: 62 threads, 3,988 comments, 2,004 distinct accounts, July 2024 to July 2026, read in two passes covering both Alignerr's own subreddit and subreddits it does not moderate. Claims required two independent accounts in two separate threads. Live listing data from our own job board, tracked from 25 July 2026. Last verified 30 July 2026.