AI Job Market Tracker 2026

AI Job Market Tracker 2026

Last updated: August 27, 2026 — Updated weekly | Maintained by: Axis Intelligence Research & David Park & Sarah Mitchell

This tracker records dated, primary-sourced signals on labour demand, wages, cohort outcomes and disclosure in an economy adopting artificial intelligence. Every entry links to a document published by the party that produced the number — a statistical agency, a payroll processor, a research institution, a job-postings platform or a regulator.

Scope boundary. Company-level workforce reductions are not logged here. A named employer cutting a named number of jobs belongs in the AI Layoff Tracker, which classifies each event by attribution tier. This tracker measures the other side of the same question: what employers are hiring for, what they pay for it, which cohorts are affected, and what they tell a regulator. No event appears in both trackers.

The headline finding: the Regulatory Disclosure Gap™

RDG™ (Regulatory Disclosure Gap) — inaugural baseline, 27 August 2026: 112,713 : 0

In the voluntary announcement channel, AI was named as a reason in 112,713 US job-cut announcements between January and July 2026. In the one mandatory regulatory channel that asks the question — New York’s WARN notice form, which since March 2025 has required employers to state whether technological innovation or automation contributed to a mass layoff — zero notices attributed layoffs to AI in the rule’s first year, across more than 160 filing companies.

RDG™ measures the distance between what employers say about AI when they are talking to investors and what they say about AI when they are filing with a labour regulator. At launch the ratio is undefined because the regulatory denominator is zero; the gap is therefore published as an absolute pair.

Metric inputs, declared. The numerator is Challenger, Gray & Christmas’s year-to-date AI-attributed announcement count for January–July 2026, taken from the firm’s July 2026 report. It is an external benchmark input, not a tracker row — the underlying announcement events are logged on the AI Layoff Tracker, and duplicating them here would double-count them across the portfolio. The denominator is entry AJMT-2026-002 below.

Two limits are stated on the record rather than buried. First, the channels cover different geographies: the numerator is national, the denominator is one state. Second, New York’s WARN threshold captures only mass layoffs at employers with 50 or more staff, so small AI-linked reductions cannot appear in it at all. Neither limit explains a gap of this size, and the metric is published with both attached. RDG™ becomes a ratio the first time the regulatory denominator is non-zero.

Relationship to AGR™. RDG™ is not a variant of the Attribution Gap Ratio™ used on the AI Layoff Tracker. AGR™ compares tiers of evidence within a set of company layoff events. RDG™ compares two disclosure channels for the same population of employers. They answer different questions and are never combined.

Sign-off note: RDG™ is a new proprietary metric and requires Editorial Director approval before the ratio form is published. The absolute-pair form is cleared for the inaugural cycle.

What the tracker currently holds

Entries	8
Date range	22 January 2026 – 12 August 2026
Signal classes represented	6
Entries carrying a verification flag	3
Entries with a fetched primary source	8 of 8
Company-level layoff events	0 — by design, see scope boundary

Entries

12 August 2026 — Stanford revision widens the early-career AI employment gap to 19%

AJMT-2026-008 · Cohort signal · Payroll administrative data · United States

The Stanford Digital Economy Lab published a revised version of Canaries in the Coal Mine?, using ADP payroll records through mid-2026. Employment among workers aged 22–25 in highly AI-exposed occupations now sits roughly 19% below where it would be had it kept pace with similarly aged workers in less-exposed occupations, widened from about 15% a year earlier. The authors attribute the movement primarily to reduced hiring rather than increased separations, and report no comparable gap among experienced workers in the same occupations. They state explicitly that they do not observe widespread economy-wide displacement, and describe the findings as descriptive indicators rather than causal estimates.

This is the tracker’s most important row, because it is the only signal in the set measured from administrative records of actual employment rather than from announcements, postings or surveys.

Primary source: Stanford Digital Economy Lab

7 August 2026 — BLS reports a 23,000 payroll decline, and no AI attribution series

AJMT-2026-007 · Context signal · Official statistics · United States 🏷 no_ai_attribution_field

Nonfarm payroll employment fell by 23,000 in July and the unemployment rate held at 4.1%, with declines in local government education and retail trade and continued growth in health care. This entry is logged for a reason that is itself the finding: the United States has no official statistic that attributes employment change to artificial intelligence. The Employment Situation contains no AI field, no AI reason code and no AI series. Every AI-attributed number in this tracker and in the RDG™ numerator therefore originates from a private data producer or a state regulator — never from national statistics.

Primary source: U.S. Bureau of Labor Statistics

6 August 2026 — Announced hiring plans reach 107,500 year to date

AJMT-2026-006 · Demand signal · Announcement aggregation · United States 🏷 not_ai_attributed

Employers announced 16,095 hiring plans in July, up 47% on June and the highest July figure since 2022. The year-to-date total of 107,500 is up 25% on the same point in 2025 and is the strongest January-to-July count since 2023. Aerospace and defence led July with 4,625 announced hires, followed by technology with 2,470 and automotive with 2,068.

The flag is the point of the entry. The hiring series carries no reason codes, so no portion of it can be assigned to AI on the evidence in the source. It is logged because the demand side of this market is routinely omitted from coverage, and because its direction currently runs opposite to the displacement narrative: the hires are concentrated in aerospace, energy and manufacturing rather than in screen work.

Primary source: Challenger, Gray & Christmas

15 July 2026 — AI-touched job titles spread beyond technology

AJMT-2026-005 · Demand signal · Job postings data · United States and Europe

Indeed Hiring Lab analysed postings from Q1 2022 to Q1 2026 across the United States, France, Germany, the Netherlands, Spain and the United Kingdom, classifying a normalised job title as AI-touched in a quarter when at least five postings under that label carry AI, GenAI, AGI or artificial intelligence in the employer’s raw title. The count of AI-touched titles rose in every market tracked. Italy was excluded on methodological grounds: “ai” is a common Italian contraction, which makes a clean keyword filter impractical.

The exclusion is recorded precisely because it is the kind of detail that separates a measurement from a headline.

Primary source: Indeed Hiring Lab

8 July 2026 — AI-exposed occupations led the postings decline, then led the recovery

AJMT-2026-004 · Mixed signal · Job postings data · United States

Comparing May 2022 with May 2026, the more AI-exposed an occupation was, the further its job postings fell — a statistically significant uncontrolled relationship, though the declines in AI-exposed vacancies began before ChatGPT’s release. Over the shorter May 2025 to May 2026 window the relationship inverts: the more exposed the occupation, the stronger the rebound. Software development postings are recovering as a share of total postings across most major economies tracked, Germany excepted.

Both directions come from the same dataset and the same authors. The entry is classed as a mixed signal for that reason, and is not filed under either demand or cohort.

Primary source: Indeed Hiring Lab

15 June 2026 — PwC puts the AI skills wage premium at 62%

AJMT-2026-003 · Wage signal · Job ads analysis · Global

PwC’s 2026 Global AI Jobs Barometer analysed more than one billion job advertisements across 27 countries and territories. Jobs requiring specific AI skills grew 69%, against 9% for the overall jobs market, and the average wage premium for AI skills rose to 62% from 57% a year earlier — as high as 118% in consumer markets and 16% in government and public sector work. Headcount growth at the most AI-exposed companies outpaced the least exposed, 52% against 36% on a 2018 baseline.

On entry level, the report cuts against the simpler version of the story: across 2.4 million US entry-level jobs analysed, roles most exposed to AI are seven times more likely to require traditionally senior human-intensive skills such as leadership and judgement. Openings for those “seniorised” entry-level roles grew 35% since 2019 while other entry-level roles shrank 10%.

Read alongside AJMT-2026-008, the two describe the same phenomenon from opposite ends: the entry-level rung is not disappearing so much as rising out of reach of the people who used to step onto it.

Primary source: PwC

18 May 2026 — No New York WARN notice attributed a layoff to AI in the rule’s first year

AJMT-2026-002 · Disclosure gap signal · Regulatory disclosure · New York State 🏷 pending_primary_nysdol_dataset

New York became the first US state to ask the question. From March 2025 its WARN notice form has required employers to indicate whether technological innovation or automation contributed to a reported layoff, and to name the technology if the box is checked. In the first year of the requirement, more than 160 companies filed WARN notices and not one attributed the layoffs to AI or automation. Filed reasons were economic circumstances, contract termination, bankruptcy, sale of the business and relocation.

This is the regulatory denominator of RDG™. The flag is honest about the sourcing: the finding is drawn from published legal analysis of the filings, and Axis Intelligence Research has not yet machine-verified it against the NYSDOL WARN dataset itself. That verification is scheduled for the next cycle and the flag stays until it clears.

Primary source: Hunton Andrews Kurth LLP

22 January 2026 — AI mentions reach a record 4.2% of US job postings

AJMT-2026-001 · Demand signal · Job postings data · United States

The share of US job postings containing at least one AI-related term rose to 4.2% in December 2025, its highest recorded level. The absolute count of AI-mentioning postings finished 2025 at 134% above February 2020 levels while remaining below the early-2022 peak — and, critically, it rose while overall postings continued to decline. Software development, IT systems and solutions, and scientific research and development each mentioned AI in 20% or more of postings. The same divergence appears in human resources, banking and finance, management, marketing, project management and accounting.

This is the tracker’s baseline demand row: the point from which the 2026 series is measured.

Primary source: Indeed Hiring Lab

How we curate this list

The definitions below are the methodology. They determine what enters the tracker and what does not, and they are the part of this page that should be cited.

What counts

An entry is logged when all of the following hold:

1. A primary document exists and has been fetched. The source must be published by the party that produced the number: the statistical agency, the payroll processor, the research institution, the regulator or the postings platform. The document is read at entry time. A row without a fetched primary source is treated as a fabricated incident.
2. The signal is dated. Publication date, filing date or release date. Where a date is inferred, the entry carries release_date_inferred and the inference basis is stated in the entry.
3. The signal measures a market, not a firm. Postings, wages, cohort employment, official statistics, regulatory disclosure. A named employer’s headcount action is out of scope regardless of how it is framed.
4. The signal is about the labour market, not about AI capability. Model launches, capability claims and productivity demos are out of scope unless the same document states a hiring plan, a wage figure or a postings measurement.
5. The number is stated in the source. Figures are transcribed, never derived, interpolated or annualised by us.

What does not count

• Company-level layoffs and hiring announcements naming a single employer. These belong to the AI Layoff Tracker, which classifies them by attribution tier. Logging them in both places would double-count them across the portfolio and produce two Axis numbers for one event.
• Other trackers, compilations and aggregator lists. This tracker does not cite, link to or draw entries from third-party layoff or hiring compilations. Where a compilation is the only available witness to an event, the event is not logged.
• Press inference. Reason attribution that appears only in coverage and not in any document from the deciding party.
• Forecasts and predictions. Executive statements about how many jobs AI will eliminate in five years are not labour market measurements.
• Vendor case studies and ROI claims presented without a headcount, a posting count or a wage figure.
• Survey sentiment presented as outcome. Employer surveys are logged under employer_survey and never merged into demand or cohort counts.

Signal classes

Each entry carries exactly one signal class, and classes are never merged in aggregation:

Class	What it measures	Error property
official_statistics	National statistical agency series	Representative; no AI attribution exists
payroll_administrative_data	Employer payroll records at scale	Large sample, non-representative firm mix
announcement_aggregation	Employer-announced plans, reason-coded	Announced ≠ executed; reason is self-declared
job_postings_data	Vacancy text and volume	Measures demand signalling, not hires
job_ads_analysis	Cross-market ad corpora	As above, plus cross-country comparability limits
regulatory_disclosure	Mandatory filings to a labour regulator	Legally consequential; threshold-limited coverage

Direction

demand_signal, wage_signal, cohort_signal, mixed_signal, disclosure_gap_signal, context_signal. Direction describes the evidence, not our interpretation of it. A single source producing evidence in both directions is classed mixed_signal rather than split into two rows.

Aggregation rules

Never sum across signal classes. Postings shares, payroll findings, wage premia and announcement totals measure different things on different populations. The tracker publishes no aggregate headcount figure, because the only honest one would be a sum of incommensurable numbers.

Verification flags are features

Flags are published, not suppressed. pending_primary_nysdol_dataset, not_ai_attributed and no_ai_attribution_field all appear in the live table and in the CSV. A flag records a limit we found and chose to state.

What we did not log this cycle

• Four company-level events — including a payments network’s 2,600-role reduction with $563m of severance, and a Bronx hospital’s replacement of 12 utilization review nurses — because they are firm-level layoff events. They are routed to the AI Layoff Tracker under the scope boundary.
• The monthly AI-attributed announcement totals, for the same reason. They are used as the declared RDG™ numerator input and are logged as events on the AI Layoff Tracker, not duplicated here.
• Several widely circulated 2026 “AI layoff” totals, because the only available source was a third-party compilation.
• Multiple executive predictions about future white-collar job losses, because no dated measurement was attached.

---

Related trackers

• AI Layoff Tracker — company-level AI-attributed workforce reductions, classified by attribution tier (AGR™)
• AI Regulation Tracker — enactment-to-enforcement lag across jurisdictions
• EU AI Act Compliance Tracker