Worth Knowing

67-82%

How often a model preferred its own writing to a person's.

Software rewards machine writing. People do the opposite.

Across 24 simulated occupations, language models preferred resumes they had written themselves over human-written ones of matched quality in 67 to 82 percent of head-to-head comparisons. A candidate whose resume came from the same model the employer screens with was 23 to 60 percent more likely to be shortlisted.

What this means. This explains the pattern I hear most often on first calls. You get through the screen, then nothing happens. An AI-written resume can clear the software and die with the human who reads it next. Write for the person who decides.

Xu, Li and Jiang, "AI Self-preferencing in Algorithmic Hiring," presented at AIES 2025, revised June 2026. A simulation of screening rather than a record of who got hired. Source

Structural, not this year's data

+10pts

Against a gap left unexplained. Study or training gains about 15.

Naming a gap recovers most of what an unexplained gap costs

163 recruiters rated 815 candidate profiles with the reason for time out of work varied. Caregiving beat an unexplained gap by roughly 10 points of interview probability, education or training by roughly 15. Saying you had given up looking cost 12 points. Length cost about half a point a month, with one exception: gaps spent on training did not decay at all.

What this means. Three things follow. Never leave a gap unlabeled, because unlabeled is the worst case and everything beats it. Never let burnout or discouragement be the words on the page, because that is the only category that tests worse than silence. And if the gap was long, put something learned inside it, because training is the one category where length stops counting against you.

D'hert, Lippens and Baert, "Not a lucky break? Why and when a career hiatus hijacks hiring chances," Labour Economics vol. 100, June 2026. Recruiters in Flanders, Belgium, rating profiles rather than live applications, so read the direction rather than the exact figures. Source

Durable finding

7.3M

Openings. Quits at 1.9 percent, near the low end of the last decade.

Openings are flat and almost nobody is quitting

Job openings, hires, quits and layoffs were all described as little changed in July. The one real movement was hiring in professional and business services, which fell by 188,000.

What this means. Low quits means low churn, and low churn means fewer seats open up in the first place. When a search is slow right now, this is usually why. That is worth knowing before you conclude your resume is the problem, though it is also worth checking whether it is.

U.S. Bureau of Labor Statistics, Job Openings and Labor Turnover Survey, July 2026, released September 1. Source

Refreshes September 29, 2026

3,462

Out of 52,881 announced cuts in August.

AI was blamed for about one August job cut in fifteen

Employers named restructuring as the reason for 16,173 of the month's announced cuts, and artificial intelligence for 3,462. Total cuts for the month were down 38 percent against August 2025, and announced hiring plans were up sharply.

What this means. The headlines are further ahead of the data than people think. Worth one caution, though: these are the reasons companies choose to publish, and restructuring is a comfortable label for a cut that a tool made possible. Treat that figure as a floor.

Challenger, Gray & Christmas, job cut report for August 2026, published September 2. Source

Refreshes early October 2026

42%

Unemployment for the same group is 5.6 percent.

Two in five recent graduates are working below their degree

Underemployment means holding a job that does not require a bachelor's degree. The Federal Reserve Bank of New York tracks it quarterly and describes conditions for this group as still challenging.

What this means. If this is you, it is a market problem with a strategy answer. It is not evidence that you chose wrong or that something is wrong with you. It does mean the first move matters more than usual, because starting off track is harder to undo than most people expect.

Federal Reserve Bank of New York, The Labor Market for Recent College Graduates, 2026:Q2. Source

Refreshes November 2026