Home/AI training job market report — August 2026 AI training job market report — August 2026
Report summary
This report analyses all 2,818 AI training and expert data roles listed across Mercor, Micro1, SuperAnnotate and Handshake AI as of 28 August 2026. It is compiled from the live listings aggregated on this site.
Two findings stand out: demand is extraordinarily fragmented — 2,234 distinct job titles across 2,818 roles — and it reaches far beyond technology, with 30 professional categories represented.
Method
Every role listed by the seven platforms was captured, deduplicated by listing and classified into
professional categories by title. Rates were recorded where the platform published one. The dataset is
a census of what these four platforms had open on the date above, not a sample.
Scope limit worth stating plainly: this covers four platforms, not
the whole market. It does not include labs recruiting directly or platforms we do not partner with, so
treat it as a large window rather than a total.
Headline figures
2,818Open roles
2,234Distinct titles
86%Title uniqueness
62%Publish a rate
28Professional categories
922Mercor
861Micro1
342Handshake AI
320SuperAnnotate
72AfterQuery
161Turing
140Ethos AI
2,818Total open
| Platform | Roles | Share | Publish a rate |
|---|
| Mercor | 922 | 32% | 100% |
| Micro1 | 861 | 32% | 0% |
| SuperAnnotate | 320 | 12% | 100% |
| Handshake AI | 342 | 8% | 100% |
Finding 1: demand is highly fragmented
Across 2,818 open roles there are 2,234 distinct titles — a uniqueness
rate of 86%. Most titles appear exactly once.
This is unusual for a job market. It reflects how the work is commissioned: a client needs a specific
expertise for a specific evaluation project, so the role is written narrowly rather than to a standard
template. The practical consequence for candidates is that browsing by category finds work that
keyword search misses.
Finding 2: this is not a technology job market
Software engineering is the largest single category, but it is a minority of the whole. The spread:
Percentages are of the 2,206 roles that classify
cleanly by title.
Finding 3: language expertise is a substantial category in its own right
298 open roles relate to language, translation and localisation — comparable to several
established professional categories. Listings name specific languages rather than asking for
"multilingual" candidates generally, which suggests demand is driven by particular client datasets
rather than by general coverage.
Finding 4: rate transparency is now the norm
62% of listings publish a rate before you apply
(2,655 of 2,818). For a contract market this is high, and it lets candidates compare
opportunities without entering a process first.
What this means for candidates
- Search by category, not by title. With 86% title uniqueness, the exact phrase you would search for probably does not exist.
- Non-technical professionals are the larger opportunity, despite the popular framing of this as tech work.
- Narrow specialisms are an advantage. Fragmented demand rewards the specific.
- Compare rates before applying. Most listings let you.
Citation and reuse
These figures may be quoted with attribution to YO AI Labs, with a link to this page. The
underlying board is public at yoailabs.com/jobs. For questions about
methodology or a data extract, contact us through the site.
Compiled 28 August 2026 by YO AI Labs,, India. Figures reflect listings on that date and change as roles open and close.
Common questions
How many AI training jobs are available?
As of 28 August 2026, 2,818 AI training and expert data roles were open across Mercor, Micro1, SuperAnnotate, Handshake AI, AfterQuery, Turing and Ethos AI combined. That covers four platforms rather than the entire market.
What professions do AI training jobs recruit?
At least 30 distinct professional categories, led by language & translation, software engineering and finance & accounting. Non-technical professions collectively outnumber technical ones.
Do AI training jobs publish their pay rates?
Most do. 62% of the 2,818 roles analysed published a rate before application, which is high for a contract market.
