Key findings

There is not enough data in this window to state findings.

Figure 1

Tasks AI training roles describe

What do AI training roles ask people to do?

Task	Share of roles read	Most often in
Review agent trajectories or tool useSmall sample	n/a	n/a
Annotate, label or tag dataSmall sample	n/a	n/a
Compare or rank model responsesSmall sample	n/a	n/a
Hold conversations or sessionsSmall sample	n/a	n/a
Create evaluation tasks, benchmarks or test casesSmall sample	n/a	n/a
Edit or rewrite textSmall sample	n/a	n/a
Write rationales for judgmentsSmall sample	n/a	n/a
Fact-check or verify accuracySmall sample	n/a	n/a
Give feedback on reasoning or explanationsSmall sample	n/a	n/a
Write ideal responses or demonstrationsSmall sample	n/a	n/a
Perform physical tasks for robot dataSmall sample	n/a	n/a
Do hands-on professional work (models, analyses, simulations)Small sample	n/a	n/a
Review others' work (QA)Small sample	n/a	n/a
Rate or score responses against guidelinesSmall sample	n/a	n/a
Record audio, video or photosSmall sample	n/a	n/a
Red-team, jailbreak or adversarial testingSmall sample	n/a	n/a
Review or debug codeSmall sample	n/a	n/a
Review for safety or policySmall sample	n/a	n/a
Solve problems step by stepSmall sample	n/a	n/a
Take part in surveys or studiesSmall sample	n/a	n/a
Transcribe audio or videoSmall sample	n/a	n/a
Translate or localiseSmall sample	n/a	n/a
Write code or testsSmall sample	n/a	n/a
Write prompts or questionsSmall sample	n/a	n/a
Write rubrics, criteria or guidelinesSmall sample	n/a	n/a

A role mentions a task when its description contains one of the task's documented patterns, anywhere in the text: a mention is not necessarily the main task, and a role often mentions several.

Hiring steps are not tasks: a description saying applications are read against a rubric, or that applying starts with an AI interview, does not count as writing rubrics or holding conversations.

Source. AITraining.jobs Research, What AI Trainers Do, live data.

Figure 2

Tasks by kind of work

Which tasks does each kind of AI training work involve?

Task
Review agent trajectories or tool use
Annotate, label or tag data
Compare or rank model responses
Hold conversations or sessions
Create evaluation tasks, benchmarks or test cases
Edit or rewrite text
Write rationales for judgments
Fact-check or verify accuracy
Give feedback on reasoning or explanations
Write ideal responses or demonstrations
Perform physical tasks for robot data
Do hands-on professional work (models, analyses, simulations)
Review others' work (QA)
Rate or score responses against guidelines
Record audio, video or photos
Red-team, jailbreak or adversarial testing
Review or debug code
Review for safety or policy
Solve problems step by step
Take part in surveys or studies
Transcribe audio or video
Translate or localise
Write code or tests
Write prompts or questions
Write rubrics, criteria or guidelines

Sample. 0 roles open on 10 Nov 2026 whose job board publishes a full description (), n/a of the 0 open roles; the boards that publish descriptions are expert networks, so these roles lean towards expert work. Kinds of work whose roles read outweigh the average of the kinds (every one with a role read does: they differ so much that each reads on its own); each cell is the share of that kind's roles read that mention the task (roles read in the column header).

A role mentions a task when its description contains one of the task's documented patterns, anywhere in the text: a mention is not necessarily the main task, and a role often mentions several.

Every kind of work with a role read is shown.

Source. AITraining.jobs Research, What AI Trainers Do, live data.

Figure 3

Advertised pay of roles mentioning each task

Which tasks come with the highest advertised pay?

Task	Median per hour	Typical pay	Against all roles read
Review agent trajectories or tool useSmall sample	n/a	n/a	n/a
Annotate, label or tag dataSmall sample	n/a	n/a	n/a
Compare or rank model responsesSmall sample	n/a	n/a	n/a
Hold conversations or sessionsSmall sample	n/a	n/a	n/a
Create evaluation tasks, benchmarks or test casesSmall sample	n/a	n/a	n/a
Edit or rewrite textSmall sample	n/a	n/a	n/a
Write rationales for judgmentsSmall sample	n/a	n/a	n/a
Fact-check or verify accuracySmall sample	n/a	n/a	n/a
Give feedback on reasoning or explanationsSmall sample	n/a	n/a	n/a
Write ideal responses or demonstrationsSmall sample	n/a	n/a	n/a
Perform physical tasks for robot dataSmall sample	n/a	n/a	n/a
Do hands-on professional work (models, analyses, simulations)Small sample	n/a	n/a	n/a
Review others' work (QA)Small sample	n/a	n/a	n/a
Rate or score responses against guidelinesSmall sample	n/a	n/a	n/a
Record audio, video or photosSmall sample	n/a	n/a	n/a
Red-team, jailbreak or adversarial testingSmall sample	n/a	n/a	n/a
Review or debug codeSmall sample	n/a	n/a	n/a
Review for safety or policySmall sample	n/a	n/a	n/a
Solve problems step by stepSmall sample	n/a	n/a	n/a
Take part in surveys or studiesSmall sample	n/a	n/a	n/a
Transcribe audio or videoSmall sample	n/a	n/a	n/a
Translate or localiseSmall sample	n/a	n/a	n/a
Write code or testsSmall sample	n/a	n/a	n/a
Write prompts or questionsSmall sample	n/a	n/a	n/a
Write rubrics, criteria or guidelinesSmall sample	n/a	n/a	n/a

A role mentions a task when its description contains one of the task's documented patterns, anywhere in the text: a mention is not necessarily the main task, and a role often mentions several.

Pay is the hourly figure a posting advertises (the middle of a stated range), counted once per distinct rate as the job board counts it. The median and typical pay are the board's: typical pay runs from the 25th percentile of what roles pay (the middle of each advertised range) to the 90th-percentile advertised ceiling, stated from 3 distinct rates. A task's median also needs rates from at least 3 platforms; others are marked and listed after the ranked tasks. 1.00x is the median of all roles read.

A role mentioning several tasks counts in each, and tasks travel with kinds of work and seniority: a pay difference is an association, not the price of the task.

Source. AITraining.jobs Research, What AI Trainers Do, live data.

Figure 4

Tasks by platform

How do the platforms that publish descriptions differ in the work they describe?

Task
Review agent trajectories or tool use
Annotate, label or tag data
Compare or rank model responses
Hold conversations or sessions
Create evaluation tasks, benchmarks or test cases
Edit or rewrite text
Write rationales for judgments
Fact-check or verify accuracy
Give feedback on reasoning or explanations
Write ideal responses or demonstrations
Perform physical tasks for robot data
Do hands-on professional work (models, analyses, simulations)
Review others' work (QA)
Rate or score responses against guidelines
Record audio, video or photos
Red-team, jailbreak or adversarial testing
Review or debug code
Review for safety or policy
Solve problems step by step
Take part in surveys or studies
Transcribe audio or video
Translate or localise
Write code or tests
Write prompts or questions
Write rubrics, criteria or guidelines

A role mentions a task when its description contains one of the task's documented patterns, anywhere in the text: a mention is not necessarily the main task, and a role often mentions several.

Listings only: a share of a platform's open roles, not of its hires or projects. Platforms write descriptions to their own template, so a task one platform's template always names counts in every one of its roles.

A finding compares platforms on one task only among those that read on their own for its share: their roles read outweigh the average of the platforms, the weight fitted on the platforms' own counts of that task.

Source. AITraining.jobs Research, What AI Trainers Do, live data.

Figure 5

How many different tasks a role describes

Do AI training roles describe one task or a mix?

Distinct tasks mentioned	Roles	Share of roles read
No task named	0	n/a
1 task	0	n/a
2 tasks	0	n/a
3 tasks	0	n/a
4 tasks	0	n/a
5 tasks	0	n/a
6 or more tasks	0	n/a

A role mentions a task when its description contains one of the task's documented patterns, anywhere in the text: a mention is not necessarily the main task, and a role often mentions several.

Out of the 25 tasks in the taxonomy. A role naming none describes its work in words no pattern reads (a project overview, a list of requirements only).

Source. AITraining.jobs Research, What AI Trainers Do, live data.

For labs, data vendors and researchers

Buy the data behind this report

The figures above are a summary. Behind them are records we collect ourselves, every day:

0 roles tracked, each with its history
Every role on 0 job boards, read since n/a (every hour since 18 Oct 2026): its pay and place, and the first and last day it was open.
0 apply clicks in 28 days
Which roles people actually apply to, by field, pay and kind of work. It happens on this site, so nobody else can count it.
39 lab and vendor hiring boards
The job boards of 25 AI labs and 14 human-data vendors, read every day: who is hiring for human data, and who is hiring to buy it.

What you can buy

The records
The rows behind any figure, with their history, not only the totals.
A regular feed
The same figures every week or month, as files.
A custom cut or benchmark
Your countries, languages or fields, or your own platform against the market.

Building AI and need the experts behind this data? (you contract them yourself) or (we deliver the expert work).

Investing in AI-data companies? Hiring radar is these records role by role, every day, as files.

Method and definitions

The public job boards of the AI training platforms this site tracks, read every hour. Full descriptions are read for the roles whose feed publishes them. Every figure is computed by one database query over the window shown, the same query for the live report and for every edition. Each figure states its sample beneath it.

Mention: A description mentions a task when it contains one of the task's documented patterns (whole words, case-insensitive, never across two lines), anywhere in the text: the duties, the requirements or a note on the project. A mention is not necessarily the main task, and a role often mentions several.
Task: One of 25 activities a description can name, from writing prompts, ideal answers, rubrics and evaluation tasks, through rating, ranking, fact-checking and red-teaming model output, to coding, annotating, transcribing, translating, recording and physical tasks for robot data. Each has a published pattern, tested on real sentences.
Not a task: Hiring steps and company boilerplate are kept out: an application read against a rubric, an AI interview, 'no screen recording' or 'professionals, not annotators' do not count as writing rubrics, holding conversations, recording or annotating.
Pay by task: The median hourly rate advertised by the roles mentioning a task. A median needs 3 roles stating a rate, and is ranked only when they come from at least 3 platforms; a median resting on fewer platforms is shown, marked, and not ranked.
Coverage: Only roles whose job board publishes a full description are read (Mercor, micro1, AfterQuery and Braintrust, about three in ten open roles); each figure states how many that is.
Open role: A role listed by a given day and still in its platform's feed that day.
Kind of work: Each role is classified from its title and description into one of the twelve kinds of AI training work this site tracks, from AI evaluation and coding to multilingual and audio work. Every kind links to its own page.
Window: Live: the last 28 complete days, in UTC. Edition: one calendar month, in UTC.

What this data cannot tell you

Platforms that recruit without a public job board are not counted. Several large ones invite workers to projects directly.
A listing is not a hire. Some platforms keep evergreen roles open for months, and some post one listing per language or country.
Descriptions are published by about three in ten open roles, all at expert networks: the figures describe those roles, lean towards expert and senior work, and say little about microtask crowds and the platforms that publish titles only.
A mention is not the main task, and a requirement counts like a duty ('experience in translation' mentions translating). Platforms write to their own templates, so a task one template always names counts in every one of its roles.
A pattern can miss a task phrased in an unusual way; nothing is inferred beyond what the text says. A pay difference between tasks is an association: tasks travel with kinds of work and seniority.

How the job feeds are read and roles are classified: site methodology. How fresh, complete and consistent the data is, checked every day: data quality and coverage.

Editions

The live report covers the last 28 complete days and changes every day. Each complete month is also published as an edition, frozen two days after the month ends, so a figure you cite from an edition never moves.

Live data (this page)Last 28 complete days, recomputed daily
October 2026 editionFreezes on first reading

Cite and reuse

Free to quote, chart and republish under CC BY 4.0 (creativecommons.org/licenses/by/4.0) with a link to this page. Live figures change daily: cite a monthly edition for numbers that will read the same later, or keep the access date.

Citation

AITraining.jobs (2026). What AI Trainers Do (live data, 14 Oct 2026 to 10 Nov 2026). https://cyberjobs.cc/research/what-ai-trainers-do (accessed 11 Nov 2026)

BibTeX

@misc{aitrainingjobs_what_ai_trainers_do_live,
  author = {{AITraining.jobs}},
  title = {What AI Trainers Do (live data, 14 Oct 2026 to 10 Nov 2026)},
  year = {2026},
  howpublished = {\url{https://cyberjobs.cc/research/what-ai-trainers-do}},
  note = {Accessed 11 Nov 2026},
}

Data: every figure as CSV (one line per value) and JSON, sent to your work email.Pay and open-roles data

More research

All 81 research reports