Skip to content

Research report ยท October 2026 edition

On-Site AI Training Work

On 31 Oct 2026, 2.2% of 4,381 open AI training roles had to be done on-site or hybrid, 95.9% were remote and 1.9% did not say; the largest group of on-site roles was in-person data collection, and most were in the United States, the United Kingdom and India.

Read the month's story: The AI Training Job Market in October 2026

Provisional figures. This month is complete, but late data can still arrive until 3 Nov 2026, when the edition is frozen for good.

Data window
1 Oct 2026 to 31 Oct 2026
31 days, UTC
Computed
1 Nov 2026, 01:11 UTC
Final once frozen
Source
The public job boards of the AI training platforms this site tracks, read every hour: each open role's title, its location text and the workplace field its board publishes.
Method
Version 2. Definitions
CC BY 4.0, how to cite

Every figure on this page as CSV and JSON, sent to your work email.

of open roles are on-site or hybrid
2%
on-site or hybrid roles open
95
of on-site roles are in the United States
73%
new on-site listings in the window
38

Key findings

  1. 1

    2.2% of the 4,381 AI training roles open on 31 Oct 2026 (95) must be done on-site or hybrid; 95.9% are remote and 1.9% do not say.

    Figure 1
  2. 2

    83 open roles state no arrangement in their title, location or workplace field, 30.1% of them at Mindrift, so the remote share is a floor.

    Figure 2
  3. 3

    Robotics & Physical-AI Data has the largest on-site share of any kind of work with enough open roles to read on its own (10%, 10 roles); 9 kinds of work have any on-site roles at all.

    Figure 1
  4. 4

    10 platforms list on-site roles; Welo Data lists the most (36, 37.9% of all on-site roles).

    Figure 2
  5. 5

    On-site roles are open in 6 countries; the United States has the most (69, 72.6%), and the place named most is the San Francisco Bay Area (50 roles).

    Figure 3
  6. 6

    Read from their titles, the largest groups of on-site roles are in-person data collection (27 roles, 28.4%) and data labelling and quality work (19, 20%).

    Figure 5
  7. 7

    Within the same kind of work, on-site roles advertise 48.8% more than remote ones on average (weighted by on-site roles, across 5 kinds of work with at least 3 priced roles on each side: Data Annotation & Labeling $34 against $20, Law, Medicine & Finance $88 against $72, Transcription, Audio & Speech $28 against $30, Multilingual & Localization $32 against $33 and Robotics & Physical-AI Data $50 against $10).

    Figure 6

Figure 1

On-site, remote or not stated, by kind of work

Which kinds of AI training work have to be done at a site rather than from home?

Kind of workOn-site or hybridRemoteNot stated
All kinds of work
2.2%
95.9%1.9%
Research Studies & SurveysSmall sample
22.2%
72.2%5.6%
Robotics & Physical-AI Data
10.0%
63.0%27.0%
Data Annotation & Labeling
8.8%
89.5%1.7%
Multilingual & Localization
2.8%
96.5%0.7%
Transcription, Audio & Speech
2.2%
95.3%2.5%
Agentic Tasks & RL Environments
1.6%
98.4%0.0%
Law, Medicine & Finance
1.0%
97.9%1.1%
Math & STEM
0.8%
98.4%0.8%
AI Evaluation & Rating (RLHF)
0.5%
95.2%4.3%
Coding & Software Evaluation
0.3%
99.3%0.3%
Red-Teaming & AI Safety
0.0%
100.0%0.0%
Writing & Content
0.0%
100.0%0.0%

Sample. 4,381 roles open at the end of 31 Oct 2026; 95 on-site or hybrid.

On-site or hybrid: the title or location names a site ('Onsite', 'On-site London', 'In-person', 'Hybrid'), or a job board that publishes a workplace field (Lever, Ashby, Mercor's own board) marks the role not remote. Remote: the title or location says remote, or the board flags the role remote.

Not stated: neither the title, the location nor a workplace field says how the work is done. Most such roles are country-level freelance listings whose full description says remote further down, so both the on-site and the remote shares are floors, not estimates of the rest.

Hybrid roles are counted with on-site ones: the listing fields do not separate them.

Rows with fewer than 32.7 open roles are marked as a small sample and not ranked: below that the average of the rows weighs more than a row's own open roles (the weight fitted on the rows' own counts).

Source. AITraining.jobs Research, On-Site AI Training Work, October 2026 edition.

Figure 2

On-site, remote or not stated, by platform

Which platforms list AI training work that has to be done in person?

PlatformOn-site or hybridRemoteNot stated
Welo Data
19.6%
80.4%0.0%
TransPerfect DataForce
43.2%
29.5%27.3%
Mercor
6.3%
93.7%0.0%
Sigma AI
13.6%
85.2%1.2%
Mindrift
5.6%
66.7%27.8%
Innodata
16.1%
83.9%0.0%
Rise Data Labs
0.9%
99.1%0.0%
TELUS Digital AI
1.5%
95.5%3.0%
Mistral AISmall sample
33.3%
66.7%0.0%
Snorkel AI Expert CommunitySmall sample
100.0%
0.0%0.0%
Invisible
0.0%
99.8%0.2%
SuperAnnotate (SME)
0.0%
100.0%0.0%
Alignerr
0.0%
100.0%0.0%
micro1
0.0%
99.7%0.3%
AfterQuery
0.0%
100.0%0.0%
Handshake AI
0.0%
100.0%0.0%
Turing
0.0%
100.0%0.0%
RWS TrainAI Community
0.0%
100.0%0.0%
Anyone AI
0.0%
100.0%0.0%
Prolific
0.0%
100.0%0.0%
DataAnnotation
0.0%
100.0%0.0%
Weekday
0.0%
93.5%6.5%
Braintrust
0.0%
100.0%0.0%
Perle Labs
0.0%
100.0%0.0%
Claru
0.0%
100.0%0.0%
Volga Partners
0.0%
100.0%0.0%
Babel Audio
0.0%
100.0%0.0%
Outlier (Scale AI)
0.0%
100.0%0.0%
Lifted (Upwork)
0.0%
93.3%6.7%
Appen
0.0%
100.0%0.0%
OpenTrain AI
0.0%
100.0%0.0%
Terac
0.0%
100.0%0.0%
Sovrano AI
0.0%
100.0%0.0%
RemoExperts
0.0%
100.0%0.0%
Vetto
0.0%
50.0%50.0%
OneForma
0.0%
94.1%5.9%
xAI (SpaceXAI)
0.0%
100.0%0.0%
Aligned Labs
0.0%
100.0%0.0%
Surge AI
0.0%
100.0%0.0%
Scale AI Human Frontier Collective
0.0%
70.0%30.0%
iMerit
0.0%
11.1%88.9%
Shaip (Collaborate by Shaip)
0.0%
100.0%0.0%
DATAmundi
0.0%
12.5%87.5%
Pasiflora AI
0.0%
100.0%0.0%
Elsevier Health
0.0%
100.0%0.0%
Cohere
0.0%
100.0%0.0%
Pila8
0.0%
100.0%0.0%
RealTrainAI
0.0%
100.0%0.0%
RemotasksSmall sample
0.0%
100.0%0.0%
SamaSmall sample
0.0%
33.3%66.7%
Fleet AISmall sample
0.0%
100.0%0.0%
AnthropicSmall sample
0.0%
100.0%0.0%
Edison ScientificSmall sample
0.0%
100.0%0.0%
EvolutionIQSmall sample
0.0%
100.0%0.0%
HeadwaySmall sample
0.0%
100.0%0.0%
Inworld AISmall sample
0.0%
100.0%0.0%
JazzX AISmall sample
0.0%
100.0%0.0%
KastleSmall sample
0.0%
100.0%0.0%
LimbicSmall sample
0.0%
100.0%0.0%
PointClickCareSmall sample
0.0%
100.0%0.0%
TrellisSmall sample
0.0%
100.0%0.0%
Vals AISmall sample
0.0%
100.0%0.0%

Sample. 4,381 roles open at the end of 31 Oct 2026, by the platform that lists them (listings, not hires).

On-site or hybrid: the title or location names a site ('Onsite', 'On-site London', 'In-person', 'Hybrid'), or a job board that publishes a workplace field (Lever, Ashby, Mercor's own board) marks the role not remote. Remote: the title or location says remote, or the board flags the role remote.

Not stated: neither the title, the location nor a workplace field says how the work is done. Most such roles are country-level freelance listings whose full description says remote further down, so both the on-site and the remote shares are floors, not estimates of the rest.

Rows with fewer than 3.1 open roles are marked as a small sample and not ranked: below that the average of the rows weighs more than a row's own open roles (the weight fitted on the rows' own counts).

Source. AITraining.jobs Research, On-Site AI Training Work, October 2026 edition.

Figure 3

Where on-site AI training roles are, by country

In which countries is in-person AI training work open?

CountryOn-site rolesShare of on-site roles
United States
69
72.6%
United Kingdom
13
13.7%
India
8
8.4%
Spain
1
1.1%
Norway
1
1.1%
Singapore
1
1.1%
Not stated
2
2.1%

Sample. 95 on-site or hybrid roles open at the end of 31 Oct 2026, by the country their location names.

Read from the location text the posting publishes. A role open at sites in two countries counts once in each, so shares can add up to more than 100%.

A city's country is taken from the location itself (the country or US state beside it, or the one country its list names). Two capital letters after a city are read as a US state. A city that cannot be matched is listed with no country.

Source. AITraining.jobs Research, On-Site AI Training Work, October 2026 edition.

Figure 4

Cities with on-site AI training roles

Which cities have the most in-person AI training work?

City or metro areaCountryOn-site rolesNaming only this city
San Francisco Bay AreaUnited States
50
20
Los AngelesUnited States
33
6
Seattle areaUnited States
31
2
New York CityUnited States
29
0
LondonUnited Kingdom
13
13
AustinUnited States
8
0
PuneIndia
6
6
Washington, D.C.United States
4
0
Hybrid - Washington DCUnited States
3
3
HyderabadIndia
2
2
In Office - San JoseUnited States
2
2
AtlantaUnited States
1
1
BarcelonaSpain
1
1
BergenNorway
1
1
HybridNot stated
1
1
ParisNot stated
1
1
San DiegoUnited States
1
1
SingaporeSingapore
1
1
CA (HybridNot stated
1
0
ChicagoUnited States
1
0
NY (HybridNot stated
1
0

Sample. 91 of the 95 on-site roles open at the end of 31 Oct 2026 name a city; the 25 cities named most.

A role that may be worked at any of several sites counts once in each city, so 'Naming only this city' counts the roles tied to that city alone.

Cities of one metropolitan area are merged: Sunnyvale, Burlingame, Santa Clara and San Francisco count as the San Francisco Bay Area; Redmond and Bellevue as the Seattle area; Brooklyn as New York City.

Source. AITraining.jobs Research, On-Site AI Training Work, October 2026 edition.

Figure 5

What on-site AI training roles are

What do the AI training roles that must be done in person involve?

Work, read from the titleOn-site rolesShare of on-site roles
In-person data collection (studies, recording sessions and their moderators)27
28.4%
Data labelling and data quality19
20.0%
Domain expert evaluation15
15.8%
Linguistic and language projects13
13.7%
Speech and voice analysis13
13.7%
Other5
5.3%
Robotics data collection and engineering3
3.2%

Sample. 95 on-site or hybrid roles open at the end of 31 Oct 2026.

Each role counts once, in the first group whose words its title contains, in the order listed in the method (robotics first, then data collection, speech and voice, labelling, linguistic, policy, expert work).

Source. AITraining.jobs Research, On-Site AI Training Work, October 2026 edition.

Figure 6

Advertised pay of on-site and remote roles, within each kind of work

Does in-person AI training work advertise more than remote work of the same kind?

Kind of workOn-site median (per hour)Remote median (per hour)On-site against remote
Data Annotation & Labeling
$34
$20+74.4%
Law, Medicine & Finance
$88
$72+21.1%
Transcription, Audio & Speech
$28
$30-6.7%
Multilingual & Localization
$32
$33-1.5%
Robotics & Physical-AI Data
$50
$10+400.0%
Agentic Tasks & RL EnvironmentsSmall sample
n/a
$50n/a
Coding & Software EvaluationSmall sample
n/a
$60n/a
Research Studies & SurveysSmall sample
n/a
$70n/a
AI Evaluation & Rating (RLHF)Small sample
n/a
$45n/a
Math & STEMSmall sample
n/a
$40n/a

Sample. Roles open at the end of 31 Oct 2026 that state an hourly rate, in the kinds of work with on-site roles.

Medians need at least 3 roles stating a rate on each side; the comparison is made within each kind of work so the mix of work is not read as a premium.

On-site roles advertising an annual salary, a fee per session or no figure are not in the medians.

Source. AITraining.jobs Research, On-Site AI Training Work, October 2026 edition.

Figure 7

On-site roles over the window, like for like

Is in-person AI training work growing or shrinking?

Work arrangementOpen on 1 Oct 2026Open on 31 Oct 2026Change
On-site or hybridSmall sample7686
+13.2%
RemoteSmall sample2,5632,429
-5.2%
Not statedSmall sample2640
+53.8%
All roles2,6652,554
-4.2%

Sample. Roles open at the start of 1 Oct 2026 and at the end of 31 Oct 2026 on the platforms tracked throughout; new listings over the whole window.

Open counts compare only platforms tracked before the window opened, so a platform added during it does not read as growth.

New listings: roles first seen in the window on a platform already tracked before that week (a platform's first week is its back catalogue).

Every row is marked as a small sample: the rows differ no more than chance makes them, so none reads on its own and none is ranked.

Source. AITraining.jobs Research, On-Site AI Training Work, October 2026 edition.

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:

  • 11,482 roles tracked, each with its history

    Every role on 62 job boards, read since 13 Jul 2026 (every hour since 18 Oct 2026): its pay and place, and the first and last day it was open.

  • 58,632 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: each open role's title, its location text and the workplace field its board publishes. 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.

On-site or hybrid
The role must be done at a named place at least part of the week: its title or location says so ('Onsite', 'On-site London', 'In-person', 'Hybrid'), or a job board that publishes a workplace field of its own (Lever, Ashby, Mercor's board) marks it not remote. Hybrid roles are counted here because the listing fields do not separate them from on-site ones.
Remote
The role's title or location says remote (or work from home, worldwide), or its board flags it remote.
Not stated
Neither the title, the location nor a workplace field says how the work is done. On most boards the stored remote flag is inferred from the word 'remote' near the top of the posting, so a missing word is not read as on-site.
City or metro area
Read from the location text: a city's country is the country or US state beside it, or the one country its list names, never guessed. Cities of one metropolitan area are merged (Sunnyvale and Burlingame count as the San Francisco Bay Area, Redmond as the Seattle area). A role open at several sites counts once in each.
Work, read from the title
Each on-site role falls in the first of these groups whose words its title contains: robotics; in-person data collection (collection, study, recording, participants, moderators); speech and voice; labelling and data quality; linguistic projects; policy and safety evaluation; domain expert work; other.
Advertised pay
The hourly figure a posting states (the middle of a stated range), compared within each kind of work. Medians need at least 3 roles stating a rate on each side.
New listing
A role that first appeared during the window on a platform already tracked before that week. A platform's first week in this dataset is its back catalogue and is not counted as new. When this site changes how it reads a platform's board (a cap lifted, a page kind or department read too, a relevance rule changed), the roles that came into view or left it from that change to the end of its week are this site's reading, not the platform's hiring: they are not new listings or removals, and changes in open roles are measured like for like across it. Each platform's report says what changed.
Open role
A role listed by a given day and still in its platform's feed that day.
Window
Live: the last 28 complete days, in UTC. Edition: one calendar month, in UTC.

What this data cannot tell you

  • Both shares are floors. When this report was built (10 Oct 2026), a check of the full descriptions of the largest group of 'not stated' roles found nearly all of them say remote further down the posting, and a few postings that require an on-site session say so only in their description.
  • Hybrid and fully on-site roles cannot be told apart from the listing fields: descriptions of some roles counted on-site ask for on-site days each week rather than every day.
  • Location text says where a role is worked or where its applicants must live, and the two are not always distinguishable; two capital letters after a city are read as a US state.
  • Platforms that recruit without a public job board are not counted. Several large ones invite workers to in-person studies directly.
  • A listing is not a hire, and some platforms post one listing per language or site, which weighs a city with many such listings more.

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.

Cite and reuse

Free to quote, chart and republish under CC BY 4.0 with a link to this page. This edition is frozen, so its figures will read the same later.

Citation

AITraining.jobs (2026). On-Site AI Training Work, October 2026 edition. https://aitraining.jobs/research/on-site-ai-training-work/2026-10

BibTeX

@misc{aitrainingjobs_on_site_ai_training_work_2026_10,
  author = {{AITraining.jobs}},
  title = {On-Site AI Training Work, October 2026 edition},
  year = {2026},
  howpublished = {\url{https://aitraining.jobs/research/on-site-ai-training-work/2026-10}},
}
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