The Ghost Job Void: How ATS and AI Have Broken the Tech Job Market
Heather Sanford has applied to more than 3,000 jobs.
She’s a marketing professional outside Austin, unemployed since her last contract ended in May 2025, and in that time she’s watched nearly $40,000 in savings disappear. Tailored resumes, custom cover letters, the whole ritual, she’s done it by the book. She’s moving in with her father soon, just to make what’s left stretch a little further.
Her story, first reported by the Wall Street Journal, isn’t really an outlier anymore. It’s closer to the median experience of a tech-adjacent job search in 2026, which says something uncomfortable about the state of things.
Here’s the part that should make you angry, though: a good chunk of the jobs she’s applying to don’t actually exist.
The Phantom Shortage
For years, executives have talked about hiring like it’s some desperate scramble for scarce talent. Not enough engineers, not enough AI-literate hires, that sort of thing.
Job boards tell a colder story. Hiring platform Greenhouse combed through its own client postings and found that roughly one in five job ads across digital job boards in the second quarter of 2026 met the definition of a “ghost job,” a listing that collected applications but produced zero real hiring activity (no interviews, no assessments, nothing), according to the WSJ’s reporting.
Some of that is just laziness, a listing nobody remembered to take down. But plenty of it is strategic. A perpetually open req signals growth to investors even mid-freeze. It keeps a resume pipeline warm for whenever budget frees up again. And it costs the company essentially nothing to leave online, which is precisely the problem. For the applicant on the other end, it costs hours, hope, and increasingly, rent money.
Recruiters will defend it, if you ask the right one. One former hiring executive told HR Brew that keeping ghost postings up lets a company move fast when someone finally quits, no scramble required. Defensible logic, in a vacuum. It’s also, as that same executive admitted, the kind of thing that makes a company look disorganized or dishonest, depending on how charitably you’re inclined to read it.
New York lawmakers weren’t feeling especially charitable. A bill passed in June requires employers with 100 or more workers to state plainly, in the posting itself, whether a role is currently open, opening later, or just a resume-farming exercise with no real fill date attached. Violators face fines up to $2,500 per posting, doubling after 30 days. Pennsylvania has similar legislation moving through its own statehouse.
It’s a start. But a state-by-state patchwork was never going to fix a problem that mostly lives on national platforms.
Machine vs. Machine
Underneath the ghost jobs sits a stranger problem: the applications that are real now mostly get read by software, not people.
Modern Applicant Tracking Systems don’t just keyword-match anymore, not the way they did a few years back. By 2026, the better systems are running semantic analysis, flagging generic AI boilerplate, and penalizing resumes that read like they were obviously written by, well, AI, according to recent coverage of how ATS platforms evolved this year. Keyword stuffing gets you flagged now, not rewarded. A little counterintuitive, if you were raised on the old advice.
Data scientist Jodi Beggs found this out the hard way. An ATS-scoring tool dinged her resume for running two pages, then docked her again for using her middle initial inconsistently across documents. Then, almost as if to prove a point, it bumped her score for the trivial act of swapping the word “percent” for a percent sign. None of that, as far as anyone can tell, has much bearing on whether she can actually do the job.
So candidates arm up, because what else is there to do. The market for the best ai resume builder tools has exploded, each one promising real-time ATS scoring and STAR-formatted bullet points engineered to slip past whatever’s doing the screening on the other end.
On the flip side of that same transaction, HR departments are buying up what vendors pitch as the best ai ats solutions and best ai recruitment platforms on the market, systems built explicitly to filter out the flood of AI-polished applications those resume builders are cranking out. Everyone’s optimizing against everyone else, and somewhere in the middle, actual qualifications get lost in the noise.
Greenhouse CEO Dan Chait has a name for the resulting spiral, as Wired reported: the “AI doom loop,” in which candidates mass-apply with AI, employers mass-reject with AI, and trust on both sides collapses toward something close to zero.
Not everyone’s surrendered to it. Toshiba’s HR lead told Wired her team still reviews every application by hand, screening on facts a language model can’t fabricate (salary expectations, notice periods, right to work). It’s boring, deliberately so. It’s also about the only approach left that isn’t already being gamed by somebody’s algorithm on the other side of the screen.
An Actual Shortage, For Comparison
Contrast all this with the skilled trades, where the labor shortage is not a euphemism for a hiring freeze.
GM CEO Mary Barra says electricians are in “huge demand” as data centers multiply across the country, and GM has sunk $250 million into trades training since 2021 trying to keep pace. Indeed’s own data, cited in Fortune, shows blue-collar data-center jobs paying a 42% premium over comparable postings elsewhere. Mike Rowe has put it more bluntly: data-center electricians are getting poached like pro athletes, for salaries north of $260,000, no degree required. A fact that stings a little if you’re carrying six figures of student debt for a resume an algorithm won’t even open.
Some of that demand is structural rather than manufactured. Oxford Economics research, covered by Newsweek, projects that the wave of baby boomer retirements between 2026 and 2029, combined with a sharp immigration slowdown, will shrink the labor force enough that the economy may not need to add any jobs at all just to keep unemployment flat.
Tech doesn’t get to borrow that excuse. Nobody’s retiring en masse out of software engineering, not yet anyway. The shortage there is manufactured, propped up by open reqs nobody intends to fill and applicant pools nobody intends to actually read.
Two algorithms are negotiating a job market that used to involve people. So far, neither side is winning, and the humans caught in between are the only ones actually losing anything.