Monthly data story
The AI Training Job Market in October 2026
AITraining.jobs research ยท Published 1 Nov 2026
Provisional figures. October 2026 is complete, but late data can still arrive until 3 Nov 2026, when its editions are frozen for good; the editions marked provisional below can still change until then.
From 1 Oct 2026 to 31 Oct 2026, open AI training roles on the platforms tracked throughout went down 3.1% (index 96.9); platforms listed 340 new roles in the last week and the advertised median stood at $50/hr.The AI Training Jobs Index, October 2026
On 31 Oct 2026, 4,381 AI training roles were open across the job boards of 62 platforms. On the platforms tracked since 1 Oct 2026, open roles went down 2.3%, from 2,665 to 2,605 like for like; Perle Labs, SuperAnnotate (SME), Turing, RemoExperts, Sovrano AI, Babel Audio, Pila8, Anyone AI, Vetto, OneForma, Claru, Pasiflora AI, Weekday, Innodata, Rise Data Labs, Volga Partners, Lifted (Upwork), Scale AI Human Frontier Collective, Anthropic, DataAnnotation, TELUS Digital AI, iMerit, Aligned Labs, Surge AI, DATAmundi, Shaip (Collaborate by Shaip), RealTrainAI, Sama, Fleet AI, Snorkel AI Expert Community, RWS TrainAI Community, Limbic, PointClickCare, Vals AI, Cohere, Mistral AI, Headway, Elsevier Health, Edison Scientific, EvolutionIQ, Inworld AI, Kastle, Trellis and JazzX AI, first tracked during the window, added 1,749. 1,638 roles were newly listed and 2,018 removed in between.AI Training Hiring Activity, October 2026
- open roles on 31 Oct 2026
- 4,381
- new listings in the window
- 1,638
- median advertised rate
- $40/hr
- of open roles state an hourly rate
- 81%
How many roles were open
From The AI Training Jobs Index, 1 Oct 2026 to 31 Oct 2026: Provisional
On the 18 platforms tracked throughout, open roles went down 3.1% like for like (about 2,520 in the week of 5 Oct 2026 and 2,717 in the week of 19 Oct 2026 as read, a count that also moved with changes in how this site reads some of their boards).Finding 1
Platforms listed about 340 new roles a week over the last week, against 432 a week in the week before (-21.3%).Finding 2
From AI Training Hiring Activity, 1 Oct 2026 to 31 Oct 2026: Provisional
Law, Medicine & Finance was the largest kind of AI training work, with 1,541 of 4,381 open roles (35.2%) on 31 Oct 2026.Finding 1
Invisible listed the most roles, 12.2% of all open roles on 31 Oct 2026; the five largest platforms together listed 41.4%.Finding 3
The full October 2026 edition of AI Training Hiring Activity
Who hired most
From AI Training Platforms Compared, 1 Oct 2026 to 31 Oct 2026: Provisional
On 31 Oct 2026, 4,381 AI training roles were open across the job boards of 62 platforms; Invisible listed the most (12.2%) and Surge AI advertised the highest median pay ($300/hr).Summary
Invisible (12.2%), SuperAnnotate (SME) (8.7%) and Alignerr (8.2%) listed the most open roles on 31 Oct 2026; the five largest of the 62 platforms held 41.4% of all open roles.Finding 1
Of the 37 platforms with at least 3 roles stating a rate, Surge AI advertised the highest median, $300/hr, and RWS TrainAI Community the lowest, $11/hr.Finding 2
The full October 2026 edition of AI Training Platforms Compared
From AI Training Vendor Health, 1 Oct 2026 to 31 Oct 2026: Provisional
From 1 Oct 2026 to 31 Oct 2026, open roles on the 18 AI training platforms tracked throughout went down 7.3% with re-key batches set aside: 2 grew and 5 shrank by 10% or more while 5 held steady, and the largest moves in roles were OpenTrain AI (-157 roles, -19.6%) and AfterQuery (+32, +16.2%).Summary
On the 18 platforms tracked throughout the window, open roles changed by -7.3% once measured like for like across the changes in how this site reads some of their boards (2,665 to 2,632 as read, -1.2%), on 1,160 new listings against 1,289 removals; 2 grew and 5 shrank by 10% or more while 5 held steady.Finding 1
On 19 Oct 2026, Mindrift re-issued 20 listings under new identifiers: the same titles left and came back at one read (24 left and 22 arrived that day; the re-issued ones are 13.7% of its removals and 8.6% of its new listings over 1 Oct 2026 to 31 Oct 2026), so they are set aside rather than read as roles filled and opened, and kept out of its 135 new listings and 49 removals.Finding 4
From AI Training Market Concentration, 1 Oct 2026 to 31 Oct 2026: Provisional
On 31 Oct 2026, AI training listings were unconcentrated overall (HHI 522 across 62 platforms); Robotics & Physical-AI Data was the most concentrated kind of work, led by Claru with 40% of its open roles.Summary
Across all AI training work, listings are unconcentrated (HHI 522 across 62 platforms); Invisible holds 12.2% of open roles.Finding 1
On the same platforms, concentration rose from 1 Oct 2026 to 31 Oct 2026 (HHI 1,368 to 1,402) and fell in 4 of 9 kinds of work; the largest move was in Agentic Tasks & RL Environments, from 3,592 to 2,287.Finding 4
The full October 2026 edition of AI Training Market Concentration
Pay
From AI Training Pay Trends, 1 Oct 2026 to 31 Oct 2026: Provisional
From the week of 28 Sep 2026 to the week of 26 Oct 2026, on the same platforms, 11 kinds of AI training work had a market rate at both ends; Red-Teaming & AI Safety moved most (+48.5%), and Red-Teaming & AI Safety paid the highest median ($74/hr).Summary
On the same platforms, between the weeks of 28 Sep 2026 and 26 Oct 2026, the advertised median for Red-Teaming & AI Safety moved up 48.5% ($50 to $74/hr), and Data Annotation & Labeling fell 17.9% ($28 to $23/hr).Finding 1
Open Law, Medicine & Finance roles changed the most, from about 617 to 1,523 a day (+146.9%), counting platforms added along the way.Finding 4
From What Raises AI Training Pay, 1 Oct 2026 to 31 Oct 2026: Provisional
Compared within each kind of work on 31 Oct 2026, a profession named in the title went with +37.6%, a US-only location rule went with +33.4%, a senior or lead title went with +21.3% and an assessment in the description with -14.5% in advertised pay.Summary
Within the same kind of work, roles whose title names a profession or discipline advertised 37.6% more than roles whose title names none, the largest premium measured; roles whose description mentions an assessment or skills test advertised 14.5% less than described roles mentioning none.Finding 1
Roles whose title names a profession or discipline advertised 37.6% more than roles whose title names none in the same kind of work, across 11 kinds (medians $65 and $35/hr overall).Finding 2
The full October 2026 edition of What Raises AI Training Pay
From The AI Training Rate Card, 1 Oct 2026 to 31 Oct 2026: Provisional
On 31 Oct 2026, for roles open to US residents, the best-paying platform advertised 5 times the rate of the worst-paying one in the typical kind of AI training work, and up to 13.5 times in Transcription, Audio & Speech ($135 at Weekday against $10 at TELUS Digital AI).Summary
Among roles open to US residents, the widest gap for the same kind of work is in Transcription, Audio & Speech: Weekday advertises a median of $135 an hour and TELUS Digital AI $10 (13.5 times).Finding 1
Across all roles, wherever they hire, the widest gap is in Law, Medicine & Finance: Surge AI advertises a median of $300 an hour and Welo Data $12, 25 times as much, partly because the two hire in different countries.Finding 3
From How AI Training Roles Advertise Pay, 1 Oct 2026 to 31 Oct 2026: Provisional
On 31 Oct 2026, 81.4% of 4,381 open AI training roles stated an hourly rate, with a median of $40/hr and 44.5% at $50/hr or more.Summary
Math & STEM roles stated a rate most often (91.5%), Robotics & Physical-AI Data roles least often (61%).Finding 2
20 of the 62 platforms with enough open roles to read on their own stated a rate on every role; Sigma AI stated one on none of its 81 roles.Finding 3
The full October 2026 edition of How AI Training Roles Advertise Pay
Where the work is open
From Where AI Training Work Is Open, 1 Oct 2026 to 31 Oct 2026: Provisional
On 31 Oct 2026, 18.9% of open AI training roles accepted applicants from any country; residents of the United States could apply to the most (2,994 of 4,381).Summary
18.9% of the 4,381 roles open on 31 Oct 2026 accepted applicants from any country; 63.0% were limited to listed countries and 18.1% excluded some.Finding 1
Red-Teaming & AI Safety was the most open to the world (37% of roles open worldwide); Robotics & Physical-AI Data the least (5%).Finding 2
The full October 2026 edition of Where AI Training Work Is Open
From On-Site AI Training Work, 1 Oct 2026 to 31 Oct 2026: Provisional
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.Summary
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.Finding 1
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.Finding 2
What applicants chose
From Applicant Demand for AI Training Work, 1 Oct 2026 to 31 Oct 2026: Provisional
From 1 Oct 2026 to 31 Oct 2026, people on AITraining.jobs made 59,228 apply clicks; per open role, interest was highest in Data Annotation & Labeling (3.7x the average role) and lowest in Research Studies & Surveys (0.15x).Summary
Law, Medicine & Finance drew the most apply clicks: 15,923 of 53,799 (29.6%), followed by Data Annotation & Labeling (17.7%) and Transcription, Audio & Speech (15.5%).Finding 1
Per open role, Data Annotation & Labeling drew the most applicant interest: 12.61 apply clicks a week, 3.7x the average role. Research Studies & Surveys drew the least: 0.52 a week (0.15x).Finding 2
The full October 2026 edition of Applicant Demand for AI Training Work
From What Draws Applicants to AI Training Roles, 1 Oct 2026 to 31 Oct 2026: Provisional
Compared within the same kind of work from 1 Oct 2026 to 31 Oct 2026, 'freelance' in the title went with 97% fewer clicks per open role, a US-only rule with 86% fewer and a long title with 73% fewer: associations among the people who use AITraining.jobs, partly the site's own ranking of roles.Summary
Within the same kind of work, roles open worldwide drew 2.3 times the apply clicks per open role of roles open only to listed countries, the strongest association measured (12 kinds of work compared).Finding 1
Roles in the top fifth of their kind of work's advertised rates drew 0.64 times the clicks per open role of their kinds' average, against 1.01 for the bottom fifth and 1.19 for roles stating no rate.Finding 4
The full October 2026 edition of What Draws Applicants to AI Training Roles
Languages and expert fields in demand
From Languages in Demand for AI Training, 1 Oct 2026 to 31 Oct 2026: Provisional
On 31 Oct 2026, open AI training roles named 95 languages; English, Spanish and French had the most openings, and Afrikaans roles advertised the highest median pay ($35/hr).Summary
English (195), Spanish (117), French (91), German (88) and Chinese (Mandarin) (78) had the most open roles on 31 Oct 2026, of 95 languages named in role titles.Finding 1
Afrikaans roles advertised the highest median pay, $35/hr, and Burmese the lowest, $8/hr, among languages with at least 3 distinct advertised rates.Finding 2
The full October 2026 edition of Languages in Demand for AI Training
From Expert Fields in Demand for AI Training, 1 Oct 2026 to 31 Oct 2026: Provisional
On 31 Oct 2026, 2,171 open AI training roles named an expert field; Software and IT, Finance and accounting and Medicine and nursing had the most openings, and Law advertised the highest median pay ($100/hr).Summary
Software and IT (395), Finance and accounting (240), Medicine and nursing (229), Engineering (non-software) (221) and Law (154) had the most open roles on 31 Oct 2026, of 21 fields named in role titles.Finding 1
Law roles advertised the highest median pay, $100/hr, and Translation and interpreting the lowest, $30/hr, among fields with at least 3 distinct advertised rates.Finding 3
The full October 2026 edition of Expert Fields in Demand for AI Training
From The Expert Scarcity Index for AI Training, 1 Oct 2026 to 31 Oct 2026: Provisional
On 31 Oct 2026, Nuclear, CBRN and defense was the hardest expert field to recruit for AI training work (scarcity index 2.78x, 1.00 = the market), German the hardest language (2.04x) and Agentic Tasks & RL Environments the hardest kind of work (1.81x); 6 groups were scarcer than the market on all three measures: applicants per opening, time to close and pay.Summary
Nuclear, CBRN and defense is the hardest expert field to staff, with a scarcity index of 2.78x (1.00 is the market): its roles draw 0.09x the market's apply clicks per open role, its listings came down 71% less often than other roles of the same age on the same platforms, and it advertises $50/hr, 25% above the market median of $40/hr.Finding 1
German is the hardest language to staff, with a scarcity index of 2.04x (1.00 is the market): its roles draw 0.17x the market's apply clicks per open role, its listings came down 48% less often than other roles of the same age on the same platforms, and it advertises $35/hr, 12% below the market median of $40/hr.Finding 2
The full October 2026 edition of The Expert Scarcity Index for AI Training
From Programming Languages in AI Training, 1 Oct 2026 to 31 Oct 2026: Provisional
Of 302 Coding & Software Evaluation roles open on 31 Oct 2026, 71.5% name a programming language, Python most often (39.1%).Summary
Python is the language named most often in open Coding & Software Evaluation roles: 118 of 302 (39.1%), in the title or the description, ahead of TypeScript (77); 22 different languages are named, and 38.1% of the roles name one in the title.Finding 1
Where a description is published, 74.5% of 259 Coding & Software Evaluation roles name a language and 43.2% name two or more; Python is named in 463 roles read.Finding 2
The full October 2026 edition of Programming Languages in AI Training
How long roles stay open
From How Long AI Training Roles Stay Open, 3 Aug 2026 to 31 Oct 2026: Provisional
Half of the AI training roles listed from 3 Aug 2026 to 31 Oct 2026 came down within 25.3 days of being listed; Writing & Content turned over fastest (median 9.2 days), and counting only removed roles would put the median at 13 days.Summary
Half of the 3,167 AI training roles listed from 3 Aug 2026 to 31 Oct 2026 came down within 25.3 days of being listed (Kaplan-Meier median), and 44.7% were still listed four weeks after going up.Finding 1
Counting only the roles already removed puts the median at 13 days; counting the 46.0% still listed too, it is 25.3 days, 95.0% longer.Finding 2
The full October 2026 edition of How Long AI Training Roles Stay Open
From When AI Training Roles Are Posted, 1 Oct 2026 to 31 Oct 2026: Provisional
From 1 Oct 2026 to 31 Oct 2026, 90.0% of dated AI training postings fell on a weekday (Tuesday the busiest day), while people applied most on Tuesday and in the 14:00 UTC hour.Summary
Tuesday and Wednesday are the busiest posting days, with 45.5% of new postings between them (an even spread would give two days 28.6%); 90.0% carry a Monday-to-Friday date.Finding 1
Postings with a stated time are busiest in the 06:00 UTC hour (9.0% of them), and 19.1% fall in the 04:00 to 08:00 UTC block.Finding 3
The full October 2026 edition of When AI Training Roles Are Posted
What roles require
From What AI Training Roles Require, 1 Oct 2026 to 31 Oct 2026: Provisional
Of 4,000 open AI training roles with a full description on 31 Oct 2026, 25.2% mention a PhD or master's degree, 16.6% a professional credential and 20.0% an assessment or test.Summary
17.3% of the 4,000 AI training roles with a full description mention a PhD, and 25.2% mention a PhD or a master's degree.Finding 1
16.6% mention a professional credential: 124 roles a medical degree and 105 a law degree or bar admission.Finding 2
The full October 2026 edition of What AI Training Roles Require
From AI Training Work Without Experience, 1 Oct 2026 to 31 Oct 2026: Provisional
On 31 Oct 2026, 24.6% of open AI training roles (1,078) asked for no expert field, level or credential in their title and advertised a median $25/hr, and 41.1% of the roles with a full description mentioned no degree, years of experience or credential, though 15.3% of those still mentioned a test.Summary
1,078 of the 4,381 AI training roles open on 31 Oct 2026 (24.6%) are generalist roles: the title names no expert field, level or credential, and the work is not Law, Medicine & Finance or Coding & Software Evaluation.Finding 1
A further 670 open roles (15.3%) are kept out only by a word such as "Expert" or "Specialist" in the title, 77.6% of them naming a language; counted in, generalist roles would be 39.9% of open roles.Finding 2
The full October 2026 edition of AI Training Work Without Experience
From How AI Training Platforms Screen Experts, 1 Oct 2026 to 31 Oct 2026: Provisional
Of the 4,000 open roles whose full description is read (61 platforms), 14.1% state a test, interview or check, most often an assessment, test or trial task (9.0%), and 16.6% name a professional credential or licence; the 1,682 application forms read on 34 platforms require a median of 10 fields, and 85.7% require a resume.Summary
Screening can be read for 61 of the 62 platforms with open roles on 31 Oct 2026: full descriptions of 4,000 roles on 61 platforms (91.3% of open roles) and application forms of 1,682 roles on 34 (38.4%); Alignerr publishes neither.Finding 1
14.1% of the 4,000 described roles state a selection step (a test, an interview or a check), most often an assessment, test or trial task (9.0%), then an interview with a person or of an unstated kind (5.6%); 0.8% state a background check or identity verification.Finding 2
The full October 2026 edition of How AI Training Platforms Screen Experts
From AI Training Job Titles, 1 Oct 2026 to 31 Oct 2026: Provisional
On 31 Oct 2026, AI training roles were most often titled "Expert or SME", "Trainer" and "Specialist"; roles titled "Lead or manager" advertised the highest median pay ($80/hr).Summary
The most common title words were Expert or SME (24.8% of open roles), Trainer (18.7% of open roles) and Specialist (18.6% of open roles).Finding 1
Roles titled "Lead or manager" advertised the highest median ($80/hr); roles titled "Rater" the lowest ($9/hr).Finding 2
The data market
From Where AI Labs Are Pushing: The Expertise Being Hired, Week by Week, 1 Oct 2026 to 31 Oct 2026: Provisional
From the week of 5 Oct 2026 to the week of 19 Oct 2026, on the 18 platforms tracked throughout, Teaching grew fastest among expert fields (+135% open roles), German among languages (+74.6%) and Red-Teaming & AI Safety among kinds of work (+36.2%).Summary
Teaching grew fastest among expert fields on the 18 platforms tracked throughout, from about 15 open roles in the week of 5 Oct 2026 to 35 in the week of 19 Oct 2026 (+135%), with 7 new listings from 4 platforms; Psychology and mental health shrank the most (-50.9%).Finding 1
German grew fastest among languages on the 18 platforms tracked throughout, from about 50 open roles in the week of 5 Oct 2026 to 87 in the week of 19 Oct 2026 (+74.6%), with 21 new listings from 8 platforms.Finding 2
The full October 2026 edition of Where AI Labs Are Pushing: The Expertise Being Hired, Week by Week
From What AI Labs Pay for Human Data, 1 Jan 2021 to 30 Oct 2026: Provisional
125 published figures on what AI labs pay for human data, from 2021 to 30 Oct 2026: lab spend, vendor bill rates against worker pay, vendors' take, RL-environment prices, prices per task and contract sizes; the largest spend figure recorded: Surge reportedly generated $1.2 billion in revenue in 2024 from work with AI labs (Surge AI, reported).Summary
US data vendors (Surge AI, Mercor, AfterQuery, Turing) and Tencent, Alibaba, ByteDance and other Chinese labs: About $500M a year of training data sold to Chinese labs (paid by the buyer; training data services, reported, 8 Aug 2026).Finding 1
Mercor: $93/hr average contracted rate; $4M+ daily payouts; oncologist roles $350 to $450 an hour (paid to the worker; expert human data (all domains), stated, 30 Oct 2026).Finding 2
The full October 2026 edition of What AI Labs Pay for Human Data
From What AI Labs Are Buying, 20 Jun 2023 to 30 Oct 2026: Provisional
To 30 Oct 2026, the evidence shows labs buying more coding (production-like environments), computer-use app and website clones and egocentric and robotics demonstration data and less generalist annotation and microtasks and static maths and verifiable coding tasks.Summary
Rising: coding (production-like environments), computer-use app and website clones, egocentric and robotics demonstration data, enterprise workflow RL environments and data, expert judgment (creative and non-professional domains), expert judgment (professional domains), in-house human data and environment teams at labs, licensed media and content, off-the-shelf and synthetic data, RL environments (all domains), voice and conversational audio and wet-lab and experimental science data.Finding 1
Enterprise workflow RL environments and data has the most evidence (8 items).Finding 3
From AI Training Data Licensing Deals, 11 Jul 2023 to 29 Oct 2026: Provisional
From 11 Jul 2023 to 29 Oct 2026, 90 content and data licensing deals between AI companies and data owners are recorded, 22 with a published price; the largest is News Corp with OpenAI ("Could be" close to $250 million over five years, per The Wall Street Journal (as reported by VentureBeat); terms not disclosed by the parties).Summary
OpenAI signed the most deals (25), ahead of Meta (14) and Google (9).Finding 3
The most frequent kind of licensor is news publisher (44 deals), then music (15) and stock media (8).Finding 4
The full October 2026 edition of AI Training Data Licensing Deals
Every edition for October 2026
The reports below published a October 2026 edition, the profiles of each kind of work and each platform among them. Each covers the calendar month and is frozen 2 days after it ends.
Reports and datasets
- Applicant demandProvisional
- Hiring activityProvisional
- Platform scorecardProvisional
- Languages in demandProvisional
- Expert fields in demandProvisional
- Pay transparencyProvisional
- Where the work is openProvisional
- Jobs indexProvisional
- Role requirementsProvisional
- ApplicationsProvisional
- Pay trendsProvisional
- Job titlesProvisional
- Pay premiumsProvisional
- Market concentrationProvisional
- Data qualityProvisional
- How long roles stay openProvisional
- When roles are postedProvisional
- What draws applicantsProvisional
- Work without experienceProvisional
- Programming languagesProvisional
- What AI trainers doProvisional
- On-site workProvisional
- How the work is paidProvisional
- Rate cardProvisional
- Human-data pricesProvisional
- Data demandProvisional
- Licensing dealsProvisional
- Data lawsuitsProvisional
- Data brokerageProvisional
- Quality and fraudProvisional
- Pay vs minimum wageProvisional
- Where labs are pushingProvisional
- Vendor healthProvisional
- Expert scarcity indexProvisional
- Real-world data collectionProvisional
- Expert screeningProvisional
- Who looks for the workProvisional
- Applicant fitProvisional
- Talent supply and demandProvisional
- Industry fundingProvisional
- Industry revenueProvisional
- Legal trackerProvisional
- Pay studiesProvisional
Kinds of work
- AI Evaluation & Rating (RLHF) marketProvisional
- Coding & Software Evaluation marketProvisional
- Math & STEM marketProvisional
- Law, Medicine & Finance marketProvisional
- Agentic Tasks & RL Environments marketProvisional
- Data Annotation & Labeling marketProvisional
- Robotics & Physical-AI Data marketProvisional
- Multilingual & Localization marketProvisional
- Transcription, Audio & Speech marketProvisional
- Writing & Content marketProvisional
- Red-Teaming & AI Safety marketProvisional
- Research Studies & Surveys marketProvisional
Platforms
- OpenTrain AI hiring dataProvisional
- Invisible hiring dataProvisional
- micro1 hiring dataProvisional
- Welo Data hiring dataProvisional
- AfterQuery hiring dataProvisional
- Mercor hiring dataProvisional
- Sigma AI hiring dataProvisional
- Braintrust hiring dataProvisional
- Alignerr hiring dataProvisional
- TransPerfect DataForce hiring dataProvisional
- Appen hiring dataProvisional
- Mindrift hiring dataProvisional
- xAI (SpaceXAI) hiring dataProvisional
- Terac hiring dataProvisional
- SuperAnnotate (SME) hiring dataProvisional
- Turing hiring dataProvisional
- Handshake AI hiring dataProvisional
- Anyone AI hiring dataProvisional
- Perle Labs hiring dataProvisional
- Claru hiring dataProvisional
- Babel Audio hiring dataProvisional
- RemoExperts hiring dataProvisional
How this story is written
- Every sentence in quotation is an edition's own: its summary, then its findings in the edition's order, the first for each of its figures, leaving out a finding whose figures are all already quoted, at most 3 for each edition and 4 editions for each section. Nothing is chosen by hand.
- Every figure links to the place in the frozen edition that states it: the table row or figure it is read from, the finding, or the edition's summary. A frozen edition never changes, so each link reads the same number later.
- The sentence setting one month against the one before uses a figure only where both months measured it the same way, and says "rose" or "fell" only when both months read the same number of platforms' job boards; otherwise it gives the two values side by side.
- A report with no edition for the month is left out, and so is a section with none.
Cite this story
AITraining.jobs (2026). The AI Training Job Market in October 2026. https://aitraining.jobs/research/monthly/2026-10
Free to quote with credit under CC BY 4.0. To cite one figure, follow its link and cite the edition it opens: each edition page gives its own citation.