TL;DR. “Layoff-proof” is the wrong goal. The useful question is how likely a role is to be cut and how quickly the person in it lands again, and the federal data treats those as two separate problems. From January 2023 to December 2025, 3.3 million U.S. workers with three or more years of tenure lost their jobs to closures, slack work or abolished positions. By January 2026, 66.1 percent were working again, but only about 49 percent of those back in full-time wage jobs were earning as much as before, down from about 62 percent two years earlier. A CEOtudent analysis joining the BLS layoff survey (JOLTS), the Displaced Workers Survey and the new BLS AI exposure categories finds three patterns. Industries that lay people off often are not the ones whose workers struggle most to get rehired. Layoffs in information rose sharply in the first half of 2026. And office and administrative support is the one occupational group with the weakest recovery numbers, the heaviest AI exposure and a projected job decline. All data here is U.S.-only, and none of it shows that AI caused any specific layoff.
This is the market-level companion to our job replaceability audit: the industry’s layoff rhythm, how the occupation fares after a layoff, and the official ten-year outlook. The CEO move is to treat it like a portfolio review. The student move is to read the caveats as carefully as the headlines.
What does the latest federal data actually measure?
Three public datasets cover the 2023-2026 restructuring period, and each measures something different. Mixing them up is how most layoff commentary goes wrong.
- JOLTS (Job Openings and Labor Turnover Survey), BLS. This is a monthly employer survey. Its “layoffs and discharges rate” is the number of involuntary separations in a month as a percentage of employment. It tells you how often an industry cuts people. It says nothing about why or what happens to them afterwards. We pulled the not seasonally adjusted monthly rates for 28 industry series from January 2023 to August 2026 through the BLS public API. August 2026 is still marked preliminary.
- Displaced Workers Survey (DWS), BLS. This is a household survey run as a supplement to the Current Population Survey every two years. The edition released on August 27, 2026 covers people aged 20 and over who lost or left jobs between January 2023 and December 2025 “because their plant or company closed down or moved, there was insufficient work for them to do, or their position or shift was abolished”. It tracks what happened to them by January 2026. This is the recovery data.
- AI exposure categories and 2025-35 employment projections, BLS. Released alongside the 2025-35 projections, this file sorts 831 detailed occupations into four relative AI exposure groups (Low, Moderate, High, Very high) and lists each one’s projected employment change.
Joined, they separate cut risk, recovery and long-run demand.
How did displaced workers actually fare from 2023 to 2025?
The DWS headline numbers are the most direct evidence on what a layoff costs a career. They are reproduced below as published.
Table 1. Outcomes for long-tenured displaced workers, 2023-2025 (verified BLS data)
| Measure | Value |
|---|---|
| Long-tenured workers displaced, Jan 2023 to Dec 2025 | 3.3 million (up 746,000 from 2021-2023) |
| All displaced workers, any tenure | 7.4 million (up from 6.3 million) |
| Reemployed in January 2026 | 66.1% (65.7% in January 2024) |
| Unemployed in January 2026 | 18.3% |
| Left the labor force | 15.7% |
| Reemployed, ages 25 to 54 | 72.9% |
| Reemployed, ages 55 to 64 | 57.3% |
| Reemployed, age 65 and older | 38.6% |
| Lost job because position or shift was abolished | 44.4% |
| Lost job to plant or company closing or move | 32.6% |
| Lost job to insufficient work | 22.9% |
| Reemployed full-timers (reporting earnings) earning as much or more than before | about 49% (about 62% in January 2024) |
Source: BLS, Worker Displacement: 2023-2025 (USDL-26-1423), news release text and tables 1, 2 and 7.
Three things stand out. First, the largest single reason for displacement was a position or shift being abolished (44.4 percent), not a company failing. That is restructuring: the employer survives but the role does not. Second, the earnings hit got worse. The reemployment rate barely moved, but the share who matched their old pay fell from about 62 percent to about 49 percent. Getting a job back and getting your career back are different outcomes. Third, age is one of the sharpest dividing lines in the release. The reemployment rate for workers aged 55 to 64 was 15.6 percentage points below the 25 to 54 group (CEOtudent calculation from the published rates).
By volume, manufacturing accounted for 19 percent of long-tenured displacements (642,000 workers), professional and business services for 16 percent and retail trade for 10 percent.
Does a high-layoff industry mean a worse career risk?
Not necessarily, and this is the most useful finding for career planning. We matched each industry’s average monthly JOLTS layoff rate over the same three years the DWS covers (2023-2025) with the DWS reemployment and unemployment shares for workers displaced from that industry. We added each industry’s 2025 hires-to-layoffs ratio as a rough gauge of how much rehiring capacity it has.
Table 2. Cut risk versus recovery, by industry (CEOtudent analysis of BLS JOLTS and BLS Displaced Workers Survey)
| Industry of lost job | Avg monthly layoff rate, 2023-25 (%) | Layoff rate, Jan-Aug 2026 vs Jan-Aug 2024 (pts) | Hires per layoff, 2025 | Displaced, 2023-25 (thousands) | Reemployed Jan 2026 (%) | Unemployed Jan 2026 (%) |
|---|---|---|---|---|---|---|
| Construction | 2.08 | -0.30 | 1.9 | 229 | 71.3 | 21.3 |
| Professional and business services | 1.85 | +0.28 | 2.2 | 524 | 60.3 | 27.1 |
| Transportation and warehousing* | 1.59 | +0.27 | 2.3 | 116 | 60.6 | 21.8 |
| Leisure and hospitality | 1.46 | +0.14 | 3.7 | 167 | 55.3 | 9.7 |
| Information | 1.18 | +0.62 | 2.0 | 108 | 76.8 | 8.1 |
| Retail trade | 1.05 | -0.05 | 3.6 | 345 | 69.1 | 16.6 |
| Other services | 0.98 | +0.14 | 3.5 | 146 | 61.7 | 17.2 |
| Manufacturing | 0.92 | -0.12 | 2.6 | 642 | 68.6 | 13.5 |
| Private educational services | 0.76 | +0.01 | 3.1 | 87 | 55.9 | 36.5 |
| Health care and social assistance | 0.69 | +0.01 | 4.5 | 280 | 77.1 | 14.5 |
| Finance and insurance | 0.48 | -0.03 | 3.8 | 229 | 57.9 | 27.1 |
| Government | 0.36 | -0.01 | 3.8 | 179 | 57.8 | 16.4 |
Notes: JOLTS rates are not seasonally adjusted. Averages are CEOtudent calculations from the BLS JOLTS monthly series, and the 2026 comparison uses the same eight months in both years to avoid seasonal distortion. “Hires per layoff” is the ratio of the 2025 average monthly hires rate to the 2025 average monthly layoffs-and-discharges rate. It counts all hires, not only rehired laid-off workers. JOLTS series covers transportation, warehousing and utilities; the DWS row covers transportation and warehousing. DWS shares do not add to 100 because the remainder left the labor force. DWS industry data now uses the 2022 Census industry classification, which BLS says is “not strictly comparable” with earlier years.*
What the table shows:
- Layoff frequency and recovery are close to unrelated. Across these 12 industries, the rank correlation between the 2023-25 layoff rate and the reemployment rate is 0.17 (CEOtudent calculation, Spearman). That is weak, and with 12 data points it is not a reliable relationship in either direction. Construction lays people off most often of any industry here, yet its displaced workers had one of the higher reemployment rates. Finance and insurance and government rarely lay people off, but fewer than six in ten of their displaced workers were working again by January 2026.
- Low cut risk can hide high landing risk. One interpretation, which the survey does not measure directly: long careers in low-churn sectors build employer-specific skills with a thinner outside market. In the finance sub-industry, 30.9 percent of displaced long-tenured workers were still unemployed in January 2026.
- Professional and business services was the weak spot on both counts. It had the second-highest layoff rate, its rate rose again into 2026, and 27.1 percent of its displaced workers were unemployed in January 2026. Within it, professional and technical services had 31.0 percent unemployed.
- Information is the 2026 signal to watch. Its average monthly layoff rate for January to August 2026 was 1.76 percent, against 1.14 percent for the same months of 2024. Five of those eight months were above the highest single month of 2024 (1.5 percent). The rate dropped back to 1.2 in July and to a preliminary 0.9 in August, so it is too early to call a trend. Workers displaced from information in 2023-25 were reemployed at 76.8 percent, a rate BLS says increased from the prior survey. The more recent 2026 layoffs are not yet in any recovery data.
What it does not show: why any of these layoffs happened. JOLTS does not ask employers about AI, and the DWS records only the three broad reasons listed above. Any claim that a given industry’s layoffs are “AI layoffs” goes beyond what this data supports.
Which occupations combine weak recovery with high AI exposure?
Industry is only half the picture, because people change industries far more easily than they change occupations. BLS itself notes that displaced workers “were not necessarily reemployed in the same industries from which they were displaced.” So the second join works at the occupation level. We grouped the 831 occupations in the BLS AI exposure file by SOC major group to match the DWS occupation groups. For each group we calculated the share of 2025 employment in High or Very high exposure occupations and the group’s projected 2025-35 employment change.
Table 3. Recovery after displacement versus AI exposure and 10-year outlook, by occupation group (CEOtudent analysis of BLS Displaced Workers Survey and BLS AI exposure categories)
| Occupation group of lost job | Displaced, 2023-25 (thousands) | Reemployed Jan 2026 (%) | Unemployed Jan 2026 (%) | 2025 employment in High or Very high AI exposure (%) | Of which Very high (%) | Projected employment change, 2025-35 (%) |
|---|---|---|---|---|---|---|
| Office and administrative support | 316 | 63.9 | 29.5 | 95.9 | 88.4 | -4.0 |
| Sales and related | 211 | 77.3 | 9.0 | 99.8 | 66.3 | -1.4 |
| Management, business and financial operations | 916 | 65.4 | 20.9 | 94.4 | 55.0 | +5.9 |
| Professional and related | 629 | 65.1 | 20.4 | 86.1 | 34.8 | +5.3 |
| Service | 302 | 65.2 | 14.0 | 11.2 | 0.1 | +5.8 |
| Installation, maintenance and repair | 134 | 78.5 | not published | 13.0 | 0.0 | +5.4 |
| Production | 352 | 65.2 | 13.0 | 10.3 | 0.3 | -0.4 |
| Construction and extraction | 188 | 59.6 | 27.1 | 2.1 | 0.0 | +5.1 |
| Transportation and material moving | 213 | 65.4 | 12.7 | 5.7 | 0.0 | +4.3 |
Notes: Exposure shares and projected change are CEOtudent aggregations of employment-weighted detailed occupations in the BLS file, mapped to SOC major groups (11 and 13; 15 to 29; 31 to 39; 41; 43; 47; 49; 51; 53). The DWS uses Census occupation groups, which follow SOC major groups closely but not perfectly. BLS does not publish values where the base is below 75,000. The -4.0 percent for office and administrative support matches the figure in the BLS projections release.
Read this table with BLS’s own warning in mind. BLS says reemployment rates by major occupational group were “little different from each other in January 2026.” The gaps in the reemployed column are mostly not statistically meaningful. With that caveat, one row still stands apart on every measure:
- Office and administrative support had the highest unemployed share among displaced workers (29.5 percent), the highest Very high exposure share (88.4 percent of employment), and the steepest projected decline. BLS projects the group to “shed 752,100 jobs over the 2025-35 decade, the most of any major occupational group.” It names the “continued integration of automation tools, including those powered by AI” as a reason. This is the one place where recovery data, exposure data and BLS’s own projection all point the same way.
- High exposure does not mean decline. Management, business and financial occupations are 94.4 percent High or Very high exposure, yet BLS projects them to grow 5.9 percent, well above the 3.5 percent all-occupation average. Sales is almost entirely exposed and still had the second-highest reemployment rate (77.3 percent). Exposure measures how much of the work AI could touch. It does not say whether the result is fewer people or more productive ones.
- Low exposure does not mean safe. Construction and extraction workers have almost no AI exposure, yet 27.1 percent of displaced workers in that group were unemployed in January 2026. Physical work protects against one risk. It does not protect against the business cycle.
What does the AI exposure file say about the next decade?
Grouping all 831 occupations by exposure category gives a picture that contradicts the popular “exposed jobs disappear” story.
Table 4. 2025-35 projections by relative AI exposure category (CEOtudent analysis of BLS AI exposure categories file)
| Relative AI exposure | Occupations | 2025 employment (millions) | Share of employment | Projected change, 2025-35 | Occupations projected to decline | Median of occupation median wages |
|---|---|---|---|---|---|---|
| Low | 213 | 30.3 | 17.8% | +2.5% | 76 | $49,120 |
| Moderate | 206 | 44.7 | 26.2% | +6.2% | 50 | $52,760 |
| High | 206 | 42.1 | 24.8% | +3.1% | 39 | $73,985 |
| Very high | 206 | 53.2 | 31.3% | +2.0% | 58 | $78,105 |
| All | 831 | 170.3 | 100% | +3.5% | 223 |
Note: BLS projects total employment to rise from 170.3 million to 176.2 million (+3.5 percent), versus 10.9 percent growth in 2015-25.
Very high exposure occupations are projected to grow the slowest, but they still grow (+2.0 percent). Of the 206 Very high occupations, 139 are projected to grow, 58 to decline, and 9 to stay flat. The difference is within the category, not between categories. Among Very high exposure occupations, the projected losers and winners look like this (BLS figures, employment in thousands):
Table 5. Same exposure, opposite outlooks: selected Very high exposure occupations (verified BLS data)
| Projected to shrink most (numeric) | Change 2025-35 | Projected to grow most (numeric) | Change 2025-35 |
|---|---|---|---|
| Office clerks, general | -156.2 (-6.0%) | Software developers | +174.7 (+10.2%) |
| Customer service representatives | -141.8 (-5.3%) | Management analysts | +109.2 (+10.1%) |
| Secretaries and administrative assistants (except legal, medical, executive) | -114.1 (-6.0%) | Computer and information systems managers | +108.1 (+15.8%) |
| Bookkeeping, accounting and auditing clerks | -85.6 (-5.6%) | Data scientists | +95.4 (+34.6%) |
| Data entry keyers | -33.6 (-25.5%) | Financial managers | +84.9 (+9.7%) |
| Payroll and timekeeping clerks | -25.4 (-15.9%) | Accountants and auditors | +79.4 (+5.0%) |
Source: BLS, AI exposure categories and 2025-35 employment projections (xlsx).
The shrinking side is mostly roles that move information through a fixed process. The growing side is mostly roles that design the process, build it, or answer for its results. Bookkeeping clerks shrink while accountants and auditors grow, even though both are rated Very high exposure. This is the same shift traced in our analysis of what compounds and what decays in career capital.
The caveats BLS insists on
The BLS methodology page lists limits that should go with every use of this file. An exposure category “is not a forecast of employment growth or decline” and “is not a worker replacement estimate”. It does “not distinguish between AI impacts from automation versus augmentation”. It is relative to other occupations rather than absolute. The theoretical sources behind it “conceptualize AI capabilities as those available no later than mid-2023”, most sources focus on language models and leave out image and video generation, and the observed-usage sources “may lean towards early adopters”. Treat the categories as a map of where AI touches work, not a list of jobs about to disappear.
How can you score your own role’s resilience?
The data points to a practical change of approach. Instead of asking whether AI will take the job, score the role on cut risk, landing risk and long-run demand separately. Then put your effort where the score is weakest. The scorecard below is a CEOtudent editorial framework. It is a structured way to read the public data above, not a validated predictive model.
The Role Resilience Scorecard (CEOtudent editorial framework)
Score each signal 0, 1 or 2. Use the tables above and the free BLS sources listed at the end.
| # | Signal | 0 points | 1 point | 2 points | Where to check |
|---|---|---|---|---|---|
| 1 | Industry cut frequency | Avg monthly layoff rate above 1.5% or rising more than 0.2 pts year on year | 0.8% to 1.5%, stable | Below 0.8%, stable | JOLTS, Table 2 |
| 2 | Industry rehiring depth | Hires per layoff below 2.5 | 2.5 to 3.5 | Above 3.5 | JOLTS, Table 2 |
| 3 | Occupation landing record | Displaced-worker unemployed share above 25% | 15% to 25% | Below 15% | DWS, Table 3 |
| 4 | Ten-year demand for your exact occupation | Projected decline | 0% to 3.5% (at or below average) | Above 3.5% | BLS projections file |
| 5 | Exposure direction | High or Very high exposure and work is mostly moving information through a fixed process | Mixed | Low exposure, or High exposure but you own judgment, design or accountability for outcomes | BLS exposure file plus your own task audit |
| 6 | Portability | Skills and network tied to one employer or one industry | Transferable within one industry | Proven in two or more industries or as an independent | Your own record |
How to read the total (0 to 12):
- 9 to 12: Compound. Your main risk is complacency. Keep the student habit going: one new adjacent skill per year, and keep your network active outside your employer.
- 5 to 8: Hedge. Fix the lowest-scoring signal first. If the weakness is signal 1 or 2 (industry), the cheapest move is often the same occupation in a different industry. If it is signal 4 or 5 (occupation), shift toward the parts of your field that are growing, following the bookkeeping-clerk-to-accountant pattern in Table 5.
- 0 to 4: Rebuild now, while employed. The DWS earnings data is the argument: only about half of reemployed full-timers matched their old pay. Moving before a layoff is usually cheaper than moving after one. For a structured way to keep options open, see optionality as a career strategy.
Three decision rules the data supports:
- Do not confuse a stable employer with a resilient career. The low-churn sectors in Table 2 (finance and insurance, government, private education) had some of the weakest reemployment outcomes. If you work in one, the CEO move is to deliberately build skills the outside market values: credentials, portfolio work, external networks.
- Inside an exposed field, move toward the decision. In every Very high exposure family in Table 5, the projected growth sits with roles that own the judgment, such as analysts, managers and auditors, rather than roles that process the inputs. Our breakdown of middle management after AI applies the same logic inside management.
- Start earlier if you are over 55. Workers aged 55 to 64 were reemployed at a rate 15.6 points below those aged 25 to 54, and workers 65 and older at 38.6 percent. People in the older groups gain the most from building portability while still employed.
What this analysis cannot tell you
The data has limits, and they belong in the reasoning, not in fine print. Everything here is U.S. data. Labor markets in Europe and elsewhere have different layoff rules and safety nets, so the levels will not transfer even if the patterns might. The DWS describes people displaced up to December 2025 and surveyed in January 2026. It cannot yet show how the information-sector layoffs of 2026 will resolve. JOLTS figures are revised, and the August 2026 value is preliminary. The joins in Tables 2 and 3 compare groups, not individuals, so they show associations that a confounder could explain (for example, the age mix of each industry). No dataset used here identifies AI as the cause of any layoff. For the entry-level side of the same market, where the pressure shows up as fewer junior openings rather than layoffs, see the entry-level squeeze.
FAQ
Which jobs are most layoff-resilient according to the data?
No occupation is layoff-proof. Health care and social assistance does best on the combination measured here: a low layoff rate (0.69 percent a month on average in 2023-25), the most hires per layoff among the matched industries (4.5), and a 77.1 percent reemployment rate for displaced workers. BLS also projects healthcare support and healthcare practitioner occupations to be the fastest-growing groups through 2035.
Are AI-exposed jobs being cut the most?
The public data cannot answer that directly, because neither JOLTS nor the Displaced Workers Survey records AI as a reason for job loss. What it shows is that office and administrative support, the most exposed major group, had the highest unemployed share among displaced workers and the steepest projected decline. Meanwhile, heavily exposed management, business and financial occupations are projected to grow faster than average.
Did layoffs increase in 2026?
Not across the economy as a whole. The average monthly layoff rate for total nonfarm employment was 1.07 percent for January to August 2026, against 1.09 percent for the same months of 2025 and 1.02 percent in 2024. Information is the exception: its rate averaged 1.76 percent over those months, up from 1.14 percent in 2024, though July and August were lower.
How long does it take to find work after a layoff?
The Displaced Workers Survey gives a status snapshot, not a duration. Of long-tenured workers displaced between 2023 and 2025, 66.1 percent were employed in January 2026, 18.3 percent were unemployed and 15.7 percent had left the labor force. Workers aged 25 to 54 did best (72.9 percent reemployed).
Does a High or Very high BLS AI exposure rating mean a job will disappear?
No. BLS states that an exposure category is not a forecast of employment decline and not a worker replacement estimate. Of the 206 Very high exposure occupations, 139 are projected to grow from 2025 to 2035. Use the rating to see which of your tasks AI touches, then check whether your specific occupation is projected to grow. For the skills employers are asking for now, see our in-demand skills analysis.
Sources
- U.S. Bureau of Labor Statistics, Job Openings and Labor Turnover Survey (JOLTS), layoffs and discharges rate and hires rate by industry, not seasonally adjusted, monthly, January 2023 to August 2026 (August 2026 preliminary), retrieved via the BLS Public Data API, October 2026.
- U.S. Bureau of Labor Statistics, “Worker Displacement: 2023-2025,” news release USDL-26-1423, August 27, 2026.
- U.S. Bureau of Labor Statistics, Displaced Workers Survey tables 4, 5 and 7 (long-tenured displaced workers by industry and occupation of lost job, and earnings of the reemployed), January 2026 Current Population Survey supplement.
- U.S. Bureau of Labor Statistics, “AI exposure categories and 2025-35 employment projections,” Employment Projections program data file, 2026.
- U.S. Bureau of Labor Statistics, “Artificial Intelligence (AI) exposure categories,” Employment Projections program methodology and limitations page, August 27, 2026.
- U.S. Bureau of Labor Statistics, “Employment Projections – 2025-2035,” news release USDL-26-1422, August 27, 2026.
This content was compiled with the support of AI following in-depth research, then written and prepared for publication by the CEOtudent editorial team.
