One in Four LinkedIn Jobs Is a Ghost. AI Didn't Just Eat the Work — It Poisoned the Signal You Hire By.
A new analysis from ResumeUp.AI looked at U.S. job listings on LinkedIn and concluded that 27.4% of them are likely "ghost jobs" — postings with no actual intention to hire anyone. More than one in four. In Los Angeles it's 30.5%. Philadelphia, 30.1%. San Francisco, 26%. If you're a developer who's applied to forty roles and heard back from three, the math just stopped being a story about you and started being a story about the listings.
And here's the accelerant: generative AI has made a fake job posting free to produce at scale. A tool that can write a plausible senior-backend-engineer JD in nine seconds can write a thousand of them before lunch — to harvest résumés, to pad a "we're growing" narrative, to scrape salary expectations, to build a talent pool for a req that doesn't exist, or just to phish. The same technology that's reshaping the work inside the job has quietly polluted the market for the job. We've been so busy arguing about whether AI writes good code that we missed it corrupting the one dataset every career decision depends on.
A Free-to-Fake Signal Is Just Noise in a Signal's Clothes
The thing nobody's naming is that the labor market runs on a signal, and the signal just got noisy in a way it's never been before. A job posting used to be a costly signal — someone wrote a req, got headcount approved, paid to list it. Cost implied intent. That's why "number of open roles" worked as a read on a company's health and a field's demand. AI collapsed the cost of producing the signal to roughly zero, and the moment a signal becomes free to fake, it stops being a signal and becomes noise wearing a signal's clothes.
This is the exact same mechanism as slopsquatting, as AI-generated SEO sludge, as the open web filling with model-written filler — a system that depended on production being expensive, getting flooded the instant production became free. We keep meeting this pattern in different costumes. The job market is just the version that hits people in the wallet, the version where a new grad sends two hundred applications into a void that was 27% fake to begin with and concludes they're worthless. They're not worthless. The instrument is broken.
If You Hire, Your Own Funnel Is Part of the Problem
If you lead engineering and you hire, this changes things on both sides of the table — and the side people forget is yours.
Your own funnel is part of the problem, even if your reqs are real. Stale postings left up "to keep the pipeline warm," evergreen reqs with no headcount behind them, roles you're 90% sure you're filling internally — every one of those reads as a ghost job to a candidate, and it spends your credibility. If you want signal back, the move is boring discipline: post real reqs, close them when they're filled, and stop using the job board as a brand-awareness channel. In a market that's 27% fake, being legibly real is a recruiting advantage.
Inbound volume is now meaningless as a quality read. AI writes the fake postings and AI writes the applications answering your real ones. Five hundred applicants used to mean something; now it means the bots found your listing. Optimize for signal over volume — referrals, work-sample tasks, anything a model can't trivially fake — because résumé count is now measuring the wrong thing in both directions.
The data you brief leadership with is contaminated. When you tell your SVP "demand for ML engineers is up, postings are surging," you may be quoting a number that's a quarter ghost. Sanity-check labor-market claims against sources with skin in the game — actual hires, actual comp data — not raw posting counts. The headline number is the most polluted one.
AI's First Move Is to Wreck the Trust a System Ran On
I do real hiring for a real team, and the part that bothers me most isn't the scams — it's what a 27% ghost rate does to trust, because trust is the thing a labor market can't function without. A market where a quarter of the offers on the shelf are fake teaches everyone to assume bad faith. Candidates stop believing postings, so they spray two hundred applications instead of writing ten good ones. Recruiters drown in AI-generated spray, so they lean harder on referrals and warm intros — which quietly rebuilds the old-boys network we spent a decade trying to dismantle with open applications. The fake jobs don't just waste time. They re-close a door that openness had pried open, and the people who lose are exactly the ones without a network: career changers, new grads, anyone whose only route in was the front door.
Here's the broader pattern, and it's the one I think is the actual 2026 story underneath all the model releases: AI's first big effect on a system is usually to wreck the trust mechanism that system quietly ran on. Stack Overflow's reputation economy. Search's link graph. The dependency registry's "this package exists" assumption. And now the job market's "a posting implies a job" assumption. Each time, the technology didn't attack the thing directly — it just made faking the trust signal free, and free-to-fake is the same as broken. The orgs and the people who come out ahead will be the ones who figure out how to re-establish costly signals in a world where the cheap ones are worthless. For hiring, that means work samples, referrals, and the deeply unsexy discipline of only posting jobs you actually intend to fill. The flashy AI question is "will it take my job." The real one, this month, is "can I still trust the board that's supposed to tell me the job exists." Right now, one time in four, no.
Sources: One Quarter of Jobs Posted Online Are Fake 'Ghost Jobs,' New Study Finds | Entrepreneur · LinkedIn 2026: Nearly 1 in 3 Jobs Are Fake | The Startup · AI Ghost Jobs Scam: How Bots Are Flooding LinkedIn to Steal Data | TechCartel · The Ghost Job Epidemic: Why 30% of 2026 Job Postings Are Fake | Fonzi