Best AI Training Platforms Compared: Live Advertised Rates, Updated Daily
This page is rebuilt every day from live job listings. The rates below are not collected once and left to rot: they are read each day from what fifteen platforms are actually advertising, and the figures carry the date they were read. Every number is a platform's own published rate, never an earnings claim.
Figures below were read from live listings on . The full history is available as a CSV.
What each platform advertises right now
| Platform | Live roles | Distinct roles | Median advertised | Middle half | Sample |
|---|---|---|---|---|---|
| Micro1 | 95 | 93 | $80/hr | $55 to $100 | 95 |
| Mercor | 220 | 216 | $75/hr | $48.60 to $100 | 220 |
| Appen | 11 | 11 | $48.30/hr | $23.63 to $61.25 | 11 |
| Alignerr | 60 | 10 | $35/hr | $35 to $91.25 | 60 |
| xAI | 23 | 23 | $35/hr | $35 to $35 | 23 |
| Welocalize | 56 | 56 | $23.75/hr | $15 to $33.33 | 56 |
Publishes no rates at all: CloudFactory, Cohere, Handshake AI, Innodata, Lionbridge, Mindrift, OneForma, RWS TrainAI, Volga Partners. That is 9 of 15 platforms, and it is worth treating as a comparison point rather than a gap: two thirds of this sector tells you what a role pays before you apply, and the rest does not.
Two things to read carefully before drawing conclusions from that table.
A median advertised rate is not a wage. It is the midpoint of the ranges a platform publishes on its own listings. What reaches your account depends on approval, caps, and which payment model your project uses, and those differ enormously between platforms. Our individual platform guides cover that.
The same role posted for eight countries is one job, not eight. The "distinct roles" column exists because some platforms inflate their apparent volume that way, and the difference between the two columns tells you which.
The same work, priced across platforms
This is the comparison most people actually want, and it is the one nobody else can produce, because it needs live listings from every platform at once rather than one company's marketing page.
General annotation, rating and evaluation
- Welocalize: $20.30/hr across 24 listings
- Appen: $48.30/hr across 11 listings
- Micro1: $60/hr across 7 listings
- Mercor: $67.50/hr across 6 listings
Software engineering
- Mercor: $85/hr across 13 listings
- Alignerr: $95/hr across 5 listings
- Micro1: $100/hr across 8 listings
Finance and accounting
- Alignerr: $90/hr across 10 listings
- Micro1: $95/hr across 3 listings
- Mercor: $117.50/hr across 30 listings
Legal
- Mercor: $91.25/hr across 6 listings
- Micro1: $110/hr across 10 listings
The pattern that holds across every one of these families: the high advertised rates attach to an existing profession, not to evaluation skill. Engineers, clinicians, lawyers and accountants are being paid for the expertise they already had. General annotation and rating work sits far below them on every platform that publishes both.
If you arrived because you saw a three-figure hourly rate, the question that matters is which of these rows your background actually places you in.
How rates have moved since we started tracking
Median advertised hourly rate, Mercor and Micro1, since tracking began.
The same figures as text, sampled weekly:
| Date | Mercor | Micro1 |
|---|---|---|
| $95/hr | $52.50/hr | |
| $90/hr | $50/hr | |
| $85/hr | $52.50/hr | |
| $85/hr | $60/hr | |
| $80/hr | $60/hr | |
| $82/hr | $70/hr | |
| $77.50/hr | $70/hr | |
| $75/hr | $80/hr |
Since tracking began, Mercor has moved from $95 to $75 an hour, down 21 per cent; Micro1 has moved from $52.50 to $80 an hour, up 52 per cent.
Only platforms with at least 30 days of tracked listings and 20 live roles appear here. Others are excluded because a median drawn from a handful of listings moves for reasons that have nothing to do with rates.
How this is put together, and what it excludes
We ingest listings from fifteen platforms daily and record what each one advertises. That record is what makes this page possible, and it is also why we can show change over time rather than a single snapshot.
Every figure is the platform's own published rate, taken from its live listings. We never publish a figure someone reported earning, on this page or anywhere else, because those cannot be verified. See our earnings disclaimer.
Medians drawn from fewer than three listings are suppressed rather than shown. A median of one listing is not a median, and small samples move for reasons that have nothing to do with rates.
Platforms that publish no rates are listed rather than hidden. Their absence from the pay column is information.
Hourly roles only. Piece-rate and per-task work is excluded from the medians, because a per-task figure and an hourly figure are not comparable and averaging them would produce a number that describes nothing.
Where to go next
Each platform has its own guide covering how it screens, how it pays, and how people lose access, built from public contributor reports rather than from marketing copy. Current openings across all fifteen platforms are on our AI evaluation job board.
Which AI training platform advertises the highest rates?
Micro1 and Mercor consistently publish the highest median advertised rates, but that reflects what they recruit for rather than generosity: their high-paying roles are physicians, attorneys, finance specialists and engineers being paid for an existing profession. For general annotation and rating work the picture is very different and much flatter.
What does general annotation work pay across platforms?
General annotation, rating and evaluation roles are advertised well below the specialist tiers on every platform that publishes both, typically in the twenties to sixties per hour against three-figure rates for professional roles. The live table on this page shows the current figure for each platform.
How often is this comparison updated?
Daily. The rates are read from live listings each day and the page carries the date they were read. Where the data is stale the page states the actual date rather than claiming freshness it does not have.
Are these figures what people actually earn?
No, and we are careful about the distinction. Every figure is the midpoint of a rate a platform publishes on its own listings. What reaches someone's account depends on approval processes, hour caps and payment models that differ by platform. We never publish reported earnings, because they cannot be verified.
Why do some platforms show no rate at all?
Roughly a third of the platforms we track publish no pay information on their listings. We list them rather than hiding them, because whether a platform tells you what a role pays before you apply is itself a meaningful difference between them.
