Glonce

AI is trapping job seekers and employers in a hiring doom loop

miscThe Glonce editorsWired🔥64Glonce rating64Impact60Novelty55Relevance82Surprise60Players35Credibility77
AI is trapping job seekers and employers in a hiring doom loop

Wired describes a hiring cycle where AI use on both sides of the job market keeps making things worse for everyone. Data scientist Jodi Beggs had her resume dinged by an online scoring tool for running two pages long and for using her middle initial inconsistently; changing "percent" to the % sign raised her score. Tools like Jobscan, which costs $30 to $50 a month, work on the premise that hiring software automatically ranks candidates in an applicant tracking system (ATS) and that only the top 10 to 20 percent get seen, so applicants tailor formatting and keywords to match what the ATS wants. Wired notes that premise is not always correct: some organizations rank automatically, others still have humans manage hiring from start to finish, and whether a company does one or the other has nothing to do with its size or how many applications it gets. It comes down to philosophy and culture instead. Regardless, applicants keep trying to game AI screening as long as they believe it exists, and one recruiter got a dozen interviews and an offer despite atrocious Jobscan scores. Daniel Chait, CEO of the ATS company Greenhouse, calls the dynamic an AI doom loop: job seekers and employers each use AI to solve their own problem in ways that make the other side's problem worse, so more AI use keeps breeding more AI use with no benefit to anyone. Kim Jones, vice president of human resources at Toshiba, says her team reviews every application by hand and that polishing materials with AI will not help a candidate get past their process, though she has noticed AI showing up unwanted in interviews, in verbose answers that follow a telltale pause. At Doist, a small, fully remote company that receives many applications for every opening, head of people Nadia Vatalidis ran a test: her team fed the job descriptions and saved application materials from already filled roles into an AI short-listing process, to see whether the people they had actually hired would have made an AI-generated cut. In two of the tested cases, the person Doist hired, someone who went on to work out well after about six months, was not on the AI short list. Job seeker James Jacobsen took a different approach after five months of searching turned his job hunt into what felt like a full-time job. Rather than only polishing his materials with Claude and ChatGPT, he built what he calls the opposite of an ATS: he had Claude comb job listings, log them, and score them against a system he devised based on role type, seniority, and salary requirements across in-person, hybrid and remote options, and to remember why he had rejected a listing if it resurfaced later. He also had Claude and ChatGPT critique his design portfolio, leading to a six-hour revamp that brought a call from a prospective employer two days later, though not an offer. Chait says candidates like Jacobsen are already asking how they could possibly do more, and that doing more of the same is not the answer; instead he suggests researching companies that are not the obvious first choices, and notes that a cover letter now stands out because Jones says she almost never sees one anymore. Chait calls this the first time he can remember both sides of hiring being unhappy at once, and says his message to job seekers is that the failure is not personal: it is the system, and it does not work well.

Key facts

  • Resume-scoring tools such as Jobscan, which costs $30 to $50 per month, work on the premise that only the top 10 to 20 percent of applicants get past an employer's AI-driven applicant tracking system.
  • Greenhouse CEO Daniel Chait calls the dynamic an AI doom loop: job seekers and employers each use AI to fix their own side of hiring in ways that make the other side's problem worse.
  • At Doist, a test that fed AI a short-listing task using already filled roles found that in two of the tested instances, the candidate Doist actually hired had been left off the AI-generated short list.
  • Toshiba's Kim Jones says her team reviews every application by hand, so AI-polished materials do not help beat their process, though AI use shows up unwanted in interviews as verbose, pause-preceded answers.
  • Job seeker James Jacobsen, five months into his search, used Claude to track, score and log job listings and to critique his portfolio, a six-hour revamp that led to one employer call but no offer.

Why it matters

The story documents a feedback loop rather than a single bad actor: job seekers deploy AI to get past what they believe are AI filters, employers deploy AI to sort the resulting flood of near-identical, AI-polished applications, and each side's fix intensifies the other side's original problem. Daniel Chait, whose company Greenhouse builds applicant tracking systems, names this directly as an AI doom loop and says it is the first time in his experience that both job seekers and employers are unhappy with hiring at the same time.

Who it affects

Job seekers such as Jodi Beggs and James Jacobsen, who spend months tailoring materials and building their own AI-driven tracking systems in place of an offer; hiring teams such as Toshiba's, where VP of human resources Kim Jones says her team reviews every application by hand instead of relying on automated screening; and small employers like the remote company Doist, whose head of people Nadia Vatalidis found that AI short-listing would have screened out people the company actually hired and was glad to have.

How to use it

Wired's reporting suggests two concrete moves rather than one product to buy. For job seekers, Jacobsen's approach shows what using AI beyond resume polishing can look like: having it track, score and remember job listings against a personal set of criteria, rather than only tailoring materials to guess at ATS keywords. For employers considering AI screening, Doist's approach is a check worth running first: test any AI short-listing system against roles you already filled, using the materials you already have, before trusting it on live candidates.

How solid is it

This is a Wired feature built on named, on-the-record interviews, Jodi Beggs, Daniel Chait of Greenhouse, Kim Jones of Toshiba and Nadia Vatalidis of Doist, plus James Jacobsen's own account, alongside the reporter's conversations with dozens of unnamed recruiters, HR managers and small business owners. It is anecdotal and qualitative: the piece itself flags that the premise behind tools like Jobscan, that only the top 10 to 20 percent of applicants are ever seen, is described as "the belief, anyway" rather than a confirmed industrywide figure, and Doist's finding rests on a small, self-run test rather than a broad study.

Risks and caveats

The article does not say how many employers or job seekers overall use AI in hiring, only individual examples, and it does not name which tool gave Beggs her resume feedback beyond noting Jobscan is the best known. Doist's test covered an unspecified but apparently small batch of already filled roles, of which two showed a mismatch between the AI short list and the actual hire; that is a demonstration of a failure mode, not a measured error rate. The piece reports that the employer call did not turn into an offer but says nothing about what happened to Jacobsen's search afterward, so whether his heavier AI-driven approach worked any better than resume polishing alone is left open.

“We've got this tragic situation where each side has a problem. They're using AI to solve their own problem, but in ways that make the problem worse. And so more AI use begets more AI use, to no one's benefit. The more it's happening, the worse it gets.”

— Daniel Chait, CEO of Greenhouse, to Wired