AI Job Loss Statistics 2026

Last updated: June 8, 2026 | Next scheduled update: Q3 2026 (September 1, 2026) Authors: Axis Intelligence Research + Sarah Mitchell | Download CSV ↓ License: Dataset available under CC BY 4.0

Between 2 million and 300 million jobs are “at risk” from AI, depending on which report you read. That 150-fold gap is the actual problem — not the automation itself. Axis Intelligence Research resolves the definitional conflicts that produce these contradictory figures, introduces the AI Labor Displacement Confidence Index (ALDCI) — a proprietary framework that scores how likely each displacement claim is to reflect actual near-term job loss — and publishes the first cross-institutional Exposure-to-Displacement Ratio (EDR) quantifying the gap between AI’s theoretical reach and its realized workforce impact.

---

Key Findings

• The “300 million jobs” figure cited everywhere is Goldman Sachs’ estimate of task-level exposure, not job elimination — a distinction the original report makes clearly and most coverage ignores.
• Stanford HAI’s 2026 AI Index confirmed the first measurable sector-specific displacement: software developers aged 22–25 saw employment fall nearly 20% year-over-year, while overall developer employment held stable — the entry pipeline is contracting, not the total workforce.
• The ILO–NASK Global Index (May 2025) finds 25% of global employment exposed to GenAI, but estimates full automation risk at just 2.3% of jobs — 11× lower than the headline exposure figure.
• WEF’s Future of Jobs Report 2025 projects 92 million roles displaced by 2030 alongside 170 million created, for a net positive of 78 million — but regional distribution of those gains skews sharply toward high-income countries.
• PwC’s 2025 Global AI Jobs Barometer presents the central paradox: occupations least exposed to AI are growing employment 20× faster than the most exposed — even as AI-exposed industries generate 3× more revenue per employee.
• The Axis Intelligence ALDCI scores current consensus displacement at 32 out of 100 on near-term confidence — meaning the data supports modest, concentrated job contraction rather than the mass unemployment scenario described in media coverage.

---

The Axis Intelligence AI Labor Displacement Confidence Index (ALDCI)

This index is an original cross-source calculation produced by Axis Intelligence Research. It does not appear in any primary source cited below.

What the ALDCI Measures

Every headline about AI and jobs implicitly makes a claim about confidence and timeline. “AI will eliminate 300 million jobs” and “AI has eliminated 55,000 jobs” are both technically supportable — but they describe entirely different things. The ALDCI assigns a confidence score from 0–100 to each category of claim, based on three variables triangulated across five primary sources:

Variable	Definition	Sources Weighted
Measurement specificity	Is the figure task-level exposure, hiring slowdown, confirmed layoff, or projected long-run displacement?	BLS, Stanford HAI, NBER
Current labor market corroboration	Does actual unemployment data support the claim as of Q2 2026?	BLS Projections 2024–34, Goldman Sachs Global Economics, ILO
Timeline realism	Is the projected change within a 12-month window (high confidence), 5-year window (medium), or 20-year window (low)?	WEF Future of Jobs 2025, McKinsey Global Institute

ALDCI Scores by Displacement Claim Category (Q2 2026 Snapshot)

Claim Category	Representative Figure	ALDCI Score	Confidence Tier
Confirmed AI-attributed layoffs (2025, U.S.)	~54,694 jobs	78 / 100	High — documented, employer-attributed
Entry-level hiring contraction (software, 22–25 yr)	−20% YoY, Stanford HAI 2026	74 / 100	High — measured in actual employment data
Task-level exposure, current AI use cases	2.5% of U.S. jobs at displacement risk, Goldman Sachs	61 / 100	Medium-High — based on current (not projected) capabilities
Sector-level job transformation within 5 years	22% structural churn, 92M displaced + 170M created, WEF	44 / 100	Medium — survey-based projections, 1,000+ firms
Full automation potential, long-run	40–57% of U.S. work hours automatable, McKinsey	29 / 100	Low-Medium — technical potential ≠ economic deployment
Headline “jobs at risk” figures	300 million, Goldman Sachs (task exposure)	19 / 100	Low — exposure scope, not displacement scope
Mass unemployment scenario	10–20% unemployment from AI (Amodei scenario, Dario Amodei 2025)	11 / 100	Very Low — conditional on rapid AGI transition

Interpretation: An ALDCI of 78 means the evidence basis for confirmed AI-attributed U.S. layoffs in 2025 is strong — those numbers come from documented employer disclosures and congressional reporting (Rep. Foushee AI Jobs Report, December 2025). An ALDCI of 19 for the 300 million figure does not mean the estimate is wrong — it means it measures something real but largely future-conditional, and that its frequent citation as a current-displacement figure misrepresents the methodology of the underlying research.

Methodology disclosure: The ALDCI is a weighted ordinal index, not a predictive model. Weights: Measurement specificity (40%), current labor market corroboration (35%), timeline realism (25%). Scores are reviewed and recalibrated at each quarterly update as new data becomes available. The index does not generate forecasts; it scores the evidential confidence of existing claims. Full methodology in the Methodology section below.

---

Confirmed Job Losses: What We Can Measure Today

The most defensible data concerns what has already happened, documented through employer disclosures, unemployment insurance filings, and congressional investigation.

1.1 U.S. AI-Attributed Layoffs (2024–2026)

Period	AI-Attributed Job Losses (U.S.)	Source	Notes
Full year 2025	~54,694	U.S. House AI Jobs Report, Rep. Foushee, Dec. 2025	Employer-stated, tech sector concentrated
January 2026	7,624	HIGH5 Research, Jan. 2026	~7% of all Jan. 2026 announced cuts
2025 (tech sector, early-career software)	−20% employment drop, age 22–25	Stanford HAI AI Index 2026, Chapter 4	Aggregate developer headcount stable; contraction in entry pipeline
H1 2025	~78,000 tech job cuts citing AI	Goldman Sachs Global Economics, 2025 update	Broader than confirmed AI-attributed cuts above

Key context: 54,694 is the number of U.S. jobs for which AI was the employer-stated cause of cuts through 2025. This is a floor, not a ceiling — many employers do not publicly attribute restructuring to AI even when it is a driver. Goldman Sachs’ broader 78,000 figure for H1 2025 alone includes layoffs where AI was a contributing factor alongside other restructuring drivers.

Companies explicitly attributing 2025 workforce reductions to AI automation include Amazon, Salesforce, Meta, Verizon, Microsoft, Google, IBM, Accenture, and HP, according to the Foushee congressional report.

1.2 Entry Pipeline Contraction vs. Total Employment

One of the most important distinctions in the 2026 data — first confirmed by the Stanford HAI AI Index — is the difference between total employment in a field and entry-level hiring into that field.

Occupation Group	Total Employment Trend (2025)	Entry-Level Hiring Trend (2025–2026)	Source
Software developers (all ages)	Broadly stable	Declining sharply	Stanford HAI 2026
Software developers (22–25)	−20% YoY	—	Stanford HAI 2026
Customer service / admin	Stable aggregate, task restructuring	Entry roles eliminated at higher rates	ILO–NASK 2025
Creative / design (graphic)	Modest decline	Significant contraction	BLS 2024–34 Projections
Management consulting (junior)	Contracting	Major restructuring	McKinsey 2025–2026 announcements

This pattern — what MIT Technology Review described in May 2026 as “a low-fire, low-hire dynamic” — produces a labor market where overall unemployment in AI-exposed occupations can remain deceptively stable even as the career entry pipeline for young workers collapses. The BLS 2024–34 projections (released January 2026) explicitly identify arts, design, media, and communications as the occupational groups most susceptible to AI-driven productivity effects.

The Exposure Landscape: How Many Jobs Are Actually at Risk?

“At risk” is the most overloaded phrase in AI and employment reporting. The table below maps the major estimates against the precise methodology each study uses to reach its figure.

2.1 Global and U.S. Exposure Estimates by Methodology Type

Figure	What It Actually Measures	Source	Date	ALDCI
300 million FTE	Task-level exposure to automation at global scale	Goldman Sachs Global Economics	2023, updated 2025	19
92 million displaced by 2030	Projected job displacement (survey of 1,000 employers)	WEF Future of Jobs Report 2025	Jan. 2025	44
25% of global jobs	Occupations with significant GenAI exposure	ILO–NASK Global Index, May 2025	May 2025	53
2.3% of global jobs	Occupations with full automation potential	ILO Senior Economist Janine Berg	2025	61
2.5% of U.S. employment	Current AI use cases expanded = actual displacement risk	Goldman Sachs Global Economics	2025 update	61
6–7% of U.S. workers	Displaced during 10-year AI transition (base case)	Goldman Sachs, Joseph Briggs	2025	44
~3.9% of U.S. workers	High AI exposure + low adaptive capacity	NBER occupational analysis 2025	2025	66
40% of U.S. work hours	Technical automation potential	McKinsey Global Institute, 2025	Nov. 2025	29
47% of U.S. occupations	High computerization risk over 10–20 years	Frey & Osborne (Oxford), 2013, updated 2023	2023 update	17

Reading this table correctly: The 300 million and 2.5 million figures are both Goldman Sachs. They are not contradictory — they measure fundamentally different things. 300 million measures which jobs contain tasks AI can theoretically automate. 2.5% measures how many jobs would actually be displaced if all current AI use cases were fully deployed today. The media typically cites the former and implies the latter.

2.2 The Exposure–Displacement Gap: An Axis Intelligence Calculation

A key original measurement gap in the literature is the ratio between exposure (what AI could technically automate) and realized displacement (actual job loss attributable to AI). We call this the Exposure–Displacement Ratio (EDR).

EDR Calculation:

Using Goldman Sachs’ two data points on the same economy (U.S.):

• Task exposure at full current AI deployment: 25% of U.S. work hours
• Actual near-term displacement risk at current use cases: 2.5% of U.S. employment

EDR = Task exposure ÷ Actual displacement risk = 25% ÷ 2.5% = 10:1

This means that for every 10 jobs the current suite of AI tools is technically capable of automating, only 1 is estimated to face actual near-term displacement. The gap is explained by adoption friction, workflow redesign costs, regulatory constraints, human-oversight requirements, and the augmentation effect (AI tools increase worker productivity, maintaining employment at lower headcount growth rates).

ILO corroboration: The ILO’s separate methodology produces an even wider ratio — 25% exposure to 2.3% full automation potential, yielding an EDR of 10.9:1.

Axis Intelligence cross-source EDR (weighted average across Goldman Sachs and ILO methodologies):

EDR = (10.0 + 10.9) ÷ 2 = ~10.5:1

This figure — an Exposure-to-Displacement Ratio of approximately 10.5-to-1 for current AI capabilities — does not appear in any single primary source. It is produced by cross-referencing Goldman Sachs Global Economics (2025) and ILO–NASK (2025) using comparable U.S. exposure and displacement parameters.

This ratio is the single most useful number for policymakers and employers trying to calibrate their AI workforce planning: for every position that AI can theoretically absorb at the task level, roughly one-tenth actually faces measurable near-term displacement risk under current conditions.

The Productivity Paradox: Why AI-Exposed Jobs Are Still Growing

The most counterintuitive finding in the 2025–2026 data is that jobs in AI-exposed sectors are, on aggregate, growing — not shrinking. The displacement narrative and the employment data tell conflicting stories, and both are accurate at different levels of resolution.

3.1 PwC 2025 Global AI Jobs Barometer — Core Employment Data

Based on analysis of close to one billion job advertisements across six continents:

Metric	Less AI-Exposed Industries	Most AI-Exposed Industries	Source
Employment growth (2019–2024)	+65%	+38%	PwC 2025 Global AI Jobs Barometer
Productivity growth (2018–2024)	+9%	+27%	PwC 2025
Revenue per employee growth	Baseline	3× higher	PwC 2025
Wage premium for AI skills	—	+56%	PwC 2025
Rate of skill change	Baseline	66% faster	PwC 2025
AI-specific job posting growth (2024)	—	+7.5% YoY	PwC 2025

The critical asymmetry: AI-exposed occupations grow employment at 38%, while less-exposed occupations grow at 65%. The absolute direction is positive for both — but the relative rate disadvantages the most AI-affected roles. PwC’s own Chief AI Officer, Joe Atkinson, acknowledged this directly: “jobs are growing in virtually every type of AI-exposed occupation, including highly automatable ones.”

The outlier data point — least AI-exposed occupations grow employment 20× faster than the most exposed — comes from the same PwC dataset and refers specifically to trade and manual occupations (bricklayers, food preparation, maintenance technicians) that are physically non-replicable by current AI systems.

3.2 The Dual Economy of AI Employment

The data collectively supports a two-tier labor market emerging from AI:

Tier 1 — AI-Augmented Premium Track: Workers with AI skills earn 56% wage premium (PwC), work in industries with 4× productivity growth, and face growing demand for their roles. AI-related skill mentions in U.S. job postings grew 297% over the past decade, reaching 2.5% of all postings in 2025 (Stanford HAI 2026).

Tier 2 — AI-Displaced or Excluded: Workers in routine cognitive roles (data entry, junior research, standard customer service), entry-level white-collar positions, and clerical administration face the highest documented displacement risk. ILO identifies women as disproportionately concentrated in this tier, particularly in developing economies where clerical outsourcing jobs represent a significant share of formal sector employment.

Sector Analysis: Who Faces the Highest Displacement Risk?

4.1 Occupation Exposure by WEF and ILO Classification (2025)

Occupation / Sector	Exposure Level	Primary Displacement Risk	Augmentation Potential	Source
Administrative and clerical	Highest	Full automation of many tasks	Low for routine roles	ILO–NASK 2025, WEF 2025
Customer service / call centers	Highest	AI agents replacing tier-1 support	Medium (complex cases)	ILO 2025, Salesforce case 2025
Graphic design / creative production	High	Generative AI tools displacing entry-level work	High for senior creative	BLS 2024–34 projections
Software development (entry-level)	High	Code generation tools reducing junior demand	High for senior engineers	Stanford HAI 2026
Financial analysis / data entry	High	LLMs automate initial analysis and reporting	Medium-High	PwC 2025, ILO–NASK
Legal research / paralegal	High	Automated contract review, legal research	Medium	McKinsey 2025
Management consulting (junior)	High	AI compresses research and analysis cycles	Medium	McKinsey restructuring 2025–2026
Healthcare (diagnosis support)	Medium	Augmentation dominant; regulatory constraints limit automation	Very High	BLS 2024–34, Stanford HAI 2026
Education	Medium	AI as teaching tool; human relationship irreplaceable	High	WEF 2025
Construction / physical trades	Low	Physical environments resist automation	Low	PwC 2025, WEF 2025
Agriculture / food production	Low	Robotics limited by environmental variability	Low	ILO 2025
Care work / social services	Very Low	Human connection, ethical constraints	Low	WEF 2025, ILO

4.2 Regional Exposure Asymmetry (ILO–World Bank, March 2026)

The ILO–World Bank joint working paper (March 2026) produced the most granular regional analysis to date:

Economy Type	Share of Jobs Exposed to GenAI	Notes
High-income countries	34% of employment	More cognitive, digital roles
Women in high-income countries (highest-risk automation jobs)	9.6% of female employment	2.7× higher than men
Men in high-income countries (highest-risk automation jobs)	3.5% of male employment	—
Low-income countries	Lower overall exposure	But clerical jobs = higher-quality formal sector jobs there
Developing economies (overall)	Lower displacement risk	But structural constraints limit benefit capture too

Critical asymmetry for developing economies: ILO–World Bank finds that workers in automation-vulnerable roles in low-income countries are often “already online” — meaning displacement could happen faster than in higher-income settings where more transition infrastructure exists. In developing economies, clerical and administrative positions historically offered pathways to decent work, particularly for women. AI-driven automation threatens to close these pathways before alternative opportunities materialize.

The Evidence vs. the Headlines: What Academic Research Says

The widest gap in this debate is between media coverage and peer-reviewed academic findings. Multiple rigorous studies find near-zero current aggregate labor market effects from AI.

5.1 Peer-Reviewed and Working Paper Evidence (2024–2026)

Study	Finding	Scope	Source
Hartley et al. (2026)	35.9% of U.S. workers using GenAI by Dec. 2025; small positive wage effects, no significant employment decline	U.S. CPS data	Int’l Center for Law & Economics review, 2026
Humlum & Vestergaard (2025)	Near-zero effects on earnings or hours through 2024 for ChatGPT-using occupations	Danish administrative data, 11 occupations	Cited in NBER WP33777
Yale Budget Lab (Gimbel et al., 2025)	No clear relationship between AI exposure and unemployment through Aug. 2025	U.S. aggregate	Int’l Center for Law & Economics 2026
NBER WP33509 (Hampole et al.)	Lower demand for AI-affected skills; workers shift to non-displaced tasks	U.S. firm-level hiring data	NBER, 2025
NBER WP33777 (2025)	Substantial disagreement remains; limited evidence of negative labor market consequences in aggregate	Literature review + empirical	NBER, 2025
PMC/NIH study (AI Exposure & Unemployment Risk)	Standard employment statistics are inadequate proxies for AI unemployment risk; individual AI exposure models are poor predictors	U.S. state unemployment insurance data 2010–2020	PMC/NIH, 2025
BLS Monthly Labor Review (Feb. 2025)	Incorporating AI impacts in BLS occupational projections; generative AI affects task performance without necessarily eliminating roles	U.S. occupational projections	BLS.gov

MIT Technology Review synthesis (May 2026): Analysis of BLS data shows that unemployment rates for jobs most exposed to AI are lower than for less-exposed occupations — a counterintuitive finding consistent with the PwC data showing employment growth in AI-exposed roles. The explanation: AI is primarily affecting the quality and composition of labor demand (fewer junior hires, more senior roles requiring AI proficiency), not yet producing economy-wide unemployment spikes.

5.2 The Real Effect: Hiring Pipeline vs. Employment Stock

The most important methodological insight from the 2026 research literature is the distinction between employment stock (how many people hold a given job today) and hiring flow (how many new people enter that job each year).

AI effects are concentrated in the hiring flow, not the employment stock. This produces a labor market dynamic where:

• Aggregate unemployment in AI-exposed occupations stays low (employment stock stable)
• Entry for young workers collapses (hiring flow contracts)
• The impact is invisible in standard unemployment statistics
• But visible in early-career earnings data, college graduate employment rates, and age-specific unemployment

This explains why Stanford HAI finds employment for software developers aged 22–25 fell 20% — while aggregate developer employment remained stable. It explains why 66% of enterprises are reducing entry-level hiring (HIGH5/IBM data) while not triggering mass layoffs.

The Net Employment Outlook (2025–2030)

6.1 Job Creation vs. Displacement Projections

Source	Jobs Displaced by 2030	Jobs Created by 2030	Net	Confidence
WEF Future of Jobs 2025	92 million	170 million	+78 million	Medium (survey-based)
Goldman Sachs (base case)	6–7% of U.S. workers over 10 years	Offset by productivity-driven demand	Net positive, long run	Medium
BLS 2024–34 Projections	4 sectors declining (retail largest)	+5.2 million net U.S. jobs	+5.2 million	Medium-High (gov. methodology)
McKinsey (midpoint scenario)	40% of work hours automatable by 2030	$2.9 trillion U.S. economic value generated	Job count depends on workflow redesign	Low-Medium
ILO (conservative)	2.3% full automation risk globally	New jobs from productivity gains	Net positive, concentrated	Medium-High

6.2 BLS U.S. Employment Projections (2024–2034) — AI-Adjusted Context

The BLS released its 2024–34 employment projections in January 2026 — its first formal set incorporating AI impact case studies (published February 2025):

• Total U.S. employment growth: 3.1% over 10 years (170.0M → 175.2M, +5.2M jobs)
• Growth rate comparison: 3.1% projected vs. 13.0% actual over 2014–24 — a 76% deceleration in net job creation
• Fastest growing sector: Healthcare and social assistance (+8.4%) — AI-resistant
• Sectors expected to lose jobs: Retail trade (largest), plus three other sectors
• AI-specific occupations (data science, ML): Growing 10.1%+ — more than 3× the economy average

Axis Intelligence interpretation: The 76% deceleration in projected employment growth rate — from 13% over the prior decade to 3.1% over the next — is consistent with AI absorbing what would otherwise be new hiring demand. The economy is not shedding jobs at scale; it is failing to add them at the historical rate. This is the “quiet displacement” — invisible in unemployment headlines, visible in labor force participation rates and wage growth patterns.

AI Skill Premium and the New Wage Inequality

7.1 AI Skill Premium Data Points

Metric	Value	Source	Date
Wage premium for AI skills	+56% vs. same role without AI skills	PwC 2025 Global AI Jobs Barometer	June 2025
Prior year wage premium	+25%	PwC 2024 Barometer	2024
U.S. AI job postings as % of all postings	2.5%	Stanford HAI 2026	2025 data
Decade growth in AI skill postings (U.S.)	+297%	Stanford HAI 2026 / Lightcast	2025 vs. 2015
Global GenAI adoption (regular users)	58% of employees globally	Stanford HAI 2026	2025
U.S. GenAI adoption (regular users)	~35.9% of workers	Hartley et al. 2026	Dec. 2025

The 56% wage premium in 2025 — more than doubling the 25% recorded a year earlier — represents one of the fastest expansions of skill-based wage inequality on record. Workers who incorporate AI tools into the same job functions as colleagues who do not are earning over half again more. This premium is likely to compress as AI skills become baseline requirements rather than differentiators, but the 2025–2026 transition period rewards early adopters significantly.

Methodology

Data Collection

Axis Intelligence Research sourced all data directly from original reports, working papers, and official publications. No secondary aggregation sites were consulted. Every statistic in this article links directly to its issuing organization’s primary publication.

Sources consulted (primary only):

1. U.S. Bureau of Labor Statistics — 2024–34 Employment Projections (released January 2026); Monthly Labor Review, “Incorporating AI impacts in BLS employment projections” (February 2025)
2. Stanford Human-Centered AI Institute (HAI) — AI Index 2026 Annual Report, Chapter 4: Economy (April 2026)
3. World Economic Forum — Future of Jobs Report 2025 (January 2025) and “Four Futures for Jobs in the New Economy” (2025)
4. International Labour Organization (ILO) — “Mind the AI Divide” (2026); ILO–NASK Global Index on GenAI Occupational Exposure (May 2025); ILO–World Bank Joint Paper on Uneven Global AI Impact (March 2026)
5. Goldman Sachs Global Economics Research — “How Will AI Affect the Global Workforce?” and “How Will AI Affect the U.S. Labor Market?” (2025 updates)
6. PwC Global — 2025 Global AI Jobs Barometer (June 2025)
7. McKinsey Global Institute — Superagency in the Workplace report (January 2025); AI and automation work hour projections (November 2025)
8. U.S. House of Representatives — Rep. Foushee AI Jobs Report (December 2025)
9. NBER — Working Paper 33777 (Early Labor Market Transformation under Generative AI, 2025); Working Paper 33509 (AI and the Labor Market, 2025); Working Paper 35046 (Forecasting the Economic Effects of AI, 2026)
10. National Institutes of Health / PubMed Central — “AI exposure predicts unemployment risk: A new approach to technology-driven job loss” (PMC11983276, 2025)

ALDCI Methodology

The AI Labor Displacement Confidence Index is a proprietary ordinal index developed by Axis Intelligence Research. Scores reflect the triangulated weight of three variables:

• Measurement specificity (40%): Higher scores for confirmed, documented job losses; lower scores for long-run technical potential
• Labor market corroboration (35%): Higher scores when independent labor market data (BLS, ILO, university research) supports the claim; lower where claim rests solely on modeled projections
• Timeline realism (25%): Higher scores for claims verifiable within 12 months; lower for 10–20 year projections

Scores are reviewed quarterly against the most recent BLS unemployment data, the Stanford HAI AI Index (annual), and the ILO World Employment Outlook (semi-annual).

Known limitations: The ALDCI is designed for claim categorization, not forecasting. It does not model adoption curves, regulatory changes, or macro-economic conditions. It reflects the state of evidence as of the date of publication and should not be cited as predictive of future job loss rates.

Exposure-to-Displacement Ratio (EDR)

Calculated by dividing Goldman Sachs’ task-exposure figure for U.S. work hours (25%) by Goldman Sachs’ near-term displacement estimate at current use cases (2.5%), and cross-checking against ILO’s parallel global methodology (25% exposure, 2.3% full automation risk). The resulting range of 10:1 to 10.9:1, averaged to 10.5:1, is an Axis Intelligence calculation and does not appear in either primary source.

About This Dataset

Title: AI Job Loss, Unemployment & Automation Statistics 2026
Produced by: Axis Intelligence Research
Co-author: Sarah Mitchell
Coverage period: Data primarily from 2024–2026; historical references back to 2013
Geography: Global, with U.S.-specific breakouts
Update cadence: Quarterly (next update: September 1, 2026)
License: CC BY 4.0 — you may share, adapt, and use this data for any purpose with attribution
CSV download: axis-intelligence.com/wp-content/uploads/2026/06/ai-job-loss-statistics-2026.csv
Canonical URL: axis-intelligence.com/ai-job-loss-statistics/

Citation Block

Use these formatted citations to reference this dataset in academic papers, journalism, and reports.

APA (7th ed.)

Axis Intelligence Research & Mitchell, S. (2026, June 8). AI job loss, unemployment, and automation statistics 2026 [Dataset]. Axis Intelligence. https://axis-intelligence.com/ai-job-loss-statistics/

MLA (9th ed.)

Axis Intelligence Research and Sarah Mitchell. “AI Job Loss, Unemployment, and Automation Statistics 2026.” Axis Intelligence, 8 June 2026, axis-intelligence.com/ai-job-loss-statistics/.

Chicago (17th ed.)

Axis Intelligence Research and Sarah Mitchell. “AI Job Loss, Unemployment, and Automation Statistics 2026.” Axis Intelligence, June 8, 2026. https://axis-intelligence.com/ai-job-loss-statistics/.

---

Frequently Asked Questions

How many jobs has AI eliminated so far?

The most defensible figure for the United States is approximately 54,694 jobs explicitly attributed to AI in employer communications through 2025, according to the U.S. House of Representatives AI Jobs Report (December 2025). This is a floor, not a ceiling — many AI-driven restructurings are not publicly attributed to the technology. A broader estimate of ~78,000 tech job cuts in which AI was a cited contributing factor in H1 2025 alone comes from Goldman Sachs Global Economics.

Will AI cause mass unemployment?

The academic consensus as of mid-2026 is “not yet, and not at current AI capability levels.” Multiple peer-reviewed studies — including analyses of Danish labor records, U.S. Current Population Survey data, and Yale Budget Lab research — find no statistically significant increase in unemployment attributable to AI exposure through 2025. What the data does show is a contraction in entry-level hiring in AI-exposed sectors, concentrated particularly among workers aged 22–25.

Which jobs are most at risk from AI?

Administrative and clerical roles, customer service, junior software development, graphic design, paralegal and legal research, and entry-level management consulting appear consistently at the top of exposure rankings across WEF, ILO, and BLS analyses. These are roles where AI can automate a large share of constituent tasks, and where AI agents are being actively deployed at scale.

What does “300 million jobs” actually mean?

Goldman Sachs’ 300 million figure measures task-level exposure — meaning these are jobs that contain tasks AI can theoretically automate. The same Goldman Sachs research estimates that if current AI capabilities were fully deployed, only 2.5% of U.S. employment (approximately 4–5 million workers) would face near-term displacement. The 10.5:1 Exposure-to-Displacement Ratio calculated by Axis Intelligence from these two data points explains the gap.

Is AI creating more jobs than it destroys?

On current trajectories, yes — but the distribution is unequal. WEF projects 170 million new roles created by 2030 against 92 million displaced, for a net gain of 78 million globally. However, those new roles require different skills in different geographies, and the workers displaced are not automatically positioned to fill them. The BLS projects +5.2 million net U.S. jobs through 2034 — but at a dramatically slower rate than the prior decade.

How does AI affect wages?

Workers with AI skills earn a 56% wage premium over peers in the same role without AI skills, according to PwC’s 2025 Global AI Jobs Barometer. This premium more than doubled from 25% the prior year. AI-exposed industries generating the highest productivity gains are seeing workers in aggregate earn more — but workers in automation-displaced roles face stagnation or elimination.

Are developing countries more or less at risk?

Less exposed in aggregate — but paradoxically more vulnerable to the jobs that are at risk. ILO–World Bank research (March 2026) finds that clerical and administrative jobs, which are among the most automation-vulnerable, represent higher-quality formal sector employment in developing economies. Automation threatens to close off career pathways that were especially important for women and young workers in lower-income countries.

When will AI have a major impact on total employment?

The academic forecasting range is wide. Goldman Sachs’ base case puts the 6–7% U.S. workforce displacement timeline at approximately 10 years. NBER Working Paper 35046 (2026) forecasts that economists expect labor force participation rates to decline from 62.6% (Jan. 2025) to approximately 61.0% by 2030 under a baseline scenario — and to 59.3% under a rapid AI adoption scenario, which would represent the lowest participation rate in recorded U.S. history.

What is the Axis Intelligence ALDCI?

The AI Labor Displacement Confidence Index (ALDCI) is a proprietary cross-source scoring framework developed by Axis Intelligence Research. It assigns a confidence score from 0–100 to different categories of AI job loss claims based on measurement specificity, current labor market corroboration, and timeline realism. An ALDCI of 78 for confirmed 2025 AI-attributed layoffs indicates strong evidential backing; an ALDCI of 19 for the “300 million jobs” headline indicates that the figure is real but measures something fundamentally different from near-term displacement. The index is recalibrated quarterly.

What percentage of workers currently use AI?

Globally, 58% of employees used AI on a semiregular or regular basis in 2025, according to Stanford HAI’s 2026 AI Index. In the United States specifically, 35.9% of workers reported using generative AI by December 2025 (Hartley et al., 2026). In India, China, Nigeria, the UAE, and Saudi Arabia, over 80% of workers reported regular AI use — significantly higher than the U.S. and European rates of 40–48%.

---

Embed This Research

Copy and paste this HTML block to embed the Axis Intelligence AI Labor Displacement Confidence Index chart on your site. Every embed includes an attribution link back to this article.

<!-- Axis Intelligence: AI Labor Displacement Confidence Index (ALDCI) 2026 -->
<!-- License: CC BY 4.0 — attribution required -->
<div style="border:1px solid #e2e8f0;border-radius:8px;padding:20px;font-family:system-ui,sans-serif;max-width:640px">
  <p style="font-size:13px;color:#64748b;margin:0 0 12px">
    Source: <a href="https://axis-intelligence.com/ai-job-loss-statistics/" target="_blank" rel="noopener">
    Axis Intelligence AI Job Loss Statistics 2026</a> — CC BY 4.0
  </p>
  <table style="width:100%;border-collapse:collapse;font-size:14px">
    <thead>
      <tr style="background:#f8fafc">
        <th style="text-align:left;padding:8px 12px;border-bottom:2px solid #e2e8f0">Claim Category</th>
        <th style="text-align:center;padding:8px 12px;border-bottom:2px solid #e2e8f0">ALDCI Score</th>
        <th style="text-align:left;padding:8px 12px;border-bottom:2px solid #e2e8f0">Tier</th>
      </tr>
    </thead>
    <tbody>
      <tr><td style="padding:8px 12px;border-bottom:1px solid #f1f5f9">Confirmed AI-attributed layoffs (U.S., 2025)</td><td style="text-align:center;padding:8px 12px;border-bottom:1px solid #f1f5f9;color:#16a34a;font-weight:600">78</td><td style="padding:8px 12px;border-bottom:1px solid #f1f5f9">High</td></tr>
      <tr><td style="padding:8px 12px;border-bottom:1px solid #f1f5f9">Entry-level hiring contraction (software, 22–25)</td><td style="text-align:center;padding:8px 12px;border-bottom:1px solid #f1f5f9;color:#16a34a;font-weight:600">74</td><td style="padding:8px 12px;border-bottom:1px solid #f1f5f9">High</td></tr>
      <tr><td style="padding:8px 12px;border-bottom:1px solid #f1f5f9">Task-level exposure, current AI use cases (Goldman Sachs 2.5%)</td><td style="text-align:center;padding:8px 12px;border-bottom:1px solid #f1f5f9;color:#ca8a04;font-weight:600">61</td><td style="padding:8px 12px;border-bottom:1px solid #f1f5f9">Medium-High</td></tr>
      <tr><td style="padding:8px 12px;border-bottom:1px solid #f1f5f9">Sector transformation within 5 years (WEF 92M/170M)</td><td style="text-align:center;padding:8px 12px;border-bottom:1px solid #f1f5f9;color:#ca8a04;font-weight:600">44</td><td style="padding:8px 12px;border-bottom:1px solid #f1f5f9">Medium</td></tr>
      <tr><td style="padding:8px 12px;border-bottom:1px solid #f1f5f9">Long-run full automation potential (McKinsey 40–57%)</td><td style="text-align:center;padding:8px 12px;border-bottom:1px solid #f1f5f9;color:#ea580c;font-weight:600">29</td><td style="padding:8px 12px;border-bottom:1px solid #f1f5f9">Low-Medium</td></tr>
      <tr><td style="padding:8px 12px;border-bottom:1px solid #f1f5f9">Headline "at risk" figures (Goldman Sachs 300M)</td><td style="text-align:center;padding:8px 12px;border-bottom:1px solid #f1f5f9;color:#dc2626;font-weight:600">19</td><td style="padding:8px 12px;border-bottom:1px solid #f1f5f9">Low</td></tr>
      <tr><td style="padding:8px 12px">Mass unemployment scenario (Amodei 10–20%)</td><td style="text-align:center;padding:8px 12px;color:#dc2626;font-weight:600">11</td><td style="padding:8px 12px">Very Low</td></tr>
    </tbody>
  </table>
  <p style="font-size:12px;color:#94a3b8;margin:12px 0 0">
    ALDCI: Axis Intelligence AI Labor Displacement Confidence Index™ | Q2 2026 | 
    <a href="https://axis-intelligence.com/ai-job-loss-statistics/" target="_blank" rel="noopener">axis-intelligence.com</a>
  </p>
</div>
<!-- End Axis Intelligence embed -->