Job seekers are pouring AI into their applications to beat AI screening tools that many employers do not actually use, and Greenhouse CEO Daniel Chait says the result is a self-reinforcing loop that makes hiring worse for everyone. Applicants pay $30 to $50 per month for services like Jobscan to tune résumés for an applicant tracking system, chasing the folk-wisdom figure that only the top 10 to 20 percent of applicants ever get read while the bottom 80 percent vanish. Employers, drowning in near-identical AI-polished submissions, respond by leaning harder on AI ranking inside their ATS. Chait calls it a doom loop, and the data from inside hiring teams suggests he is not exaggerating.
The premise driving the optimization industry is that hiring committees run every résumé through automated ranking before a human sees it. That happens in some organizations. In many others, hiring is still managed personally from first review to offer, and no amount of keyword tuning changes the outcome. Chait notes that no two ATS platforms behave the same way, that the AI features inside them change from month to month, and that whether AI is used at all depends on which modules a given team has paid for and switched on.
The gap between the folklore and the reality is where the money is being spent. One recruiter documented getting a dozen interviews and an offer despite atrocious Jobscan scores. The vendors selling optimization tools charge a recurring subscription and have no incentive to move a candidate off the job market quickly.
“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
Key facts
- 01Jobscan and similar résumé-optimization tools charge $30 to $50 per month to help applicants tune materials for ATS keyword ranking.
- 02The prevailing folklore holds that only the top 10 to 20 percent of applicants clear the ATS, leaving 80 percent unread.
- 03Doist tested ATS ranking against its own past hires and found the eventual hire missed the AI-generated short list in two separate roles.
- 04James Jacobsen spent five months job searching, built a Claude-powered tracker, and got a callback two days after a six-hour AI-guided portfolio revamp — but no offer.
- 05Greenhouse CEO Daniel Chait says both job seekers and employers are unhappy, calling the AI-on-AI dynamic a doom loop.
Kim Jones, vice president of human resources at Toshiba, says her team reviews every application by hand. She has no objection to candidates using AI to polish their materials, but the polishing does not affect whether an application clears her funnel — job requirements, salary expectations, and rehire eligibility do. Where Jones has started noticing AI is in live interviews. She describes hearing a pause, then typing, then a verbose answer that does not sound like the candidate.
Doist, a fully remote company that hires internationally and receives large application volumes, ran a direct test on the ATS-ranking premise. Nadia Vatalidis, head of people at Doist, fed job descriptions and archived applicant materials from already-filled roles into an AI ranking system to see whether it would surface the people her team had actually interviewed and hired.
The AI did overlap with the human short list at the interview stage, but the eventual hires — employees who were performing well after about six months — did not consistently appear in the AI-generated top tier. In two separate roles the person Doist ended up hiring was not in the machine's short list at all. That is precisely the failure mode the optimization industry cannot see: candidates who tune for the machine may be tuning against the humans who actually make offers.
James Jacobsen, a design professional, started his job search five months ago and treated it like a full-time role. He initially used Claude and ChatGPT to align his résumé and cover letters with job descriptions, saw no traction, and pivoted. He built what amounts to an inverse ATS: a Claude-driven tracker that combs listings, scores them against a system weighted for role type, seniority, and salary by work arrangement, and logs why he passed on any role so a later reposting does not waste his time.
He then asked Claude and ChatGPT to critique his portfolio and spent six hours rebuilding it on their notes. Two days later, an employer called. No offer followed. Jacobsen has produced one of the more sophisticated individual applications of AI to a job search, and the return remains a callback.
Chait's advice to candidates cuts against the entire optimization industry. Volume is not the answer, he argues — spray-and-pray with AI-tuned résumés is precisely what got the market here. Jones adds that cover letters have become so rare that submitting one is now a differentiator, and networking still outperforms most application funnels. Chait's closing line to job seekers is that the failure is structural: it is not the candidate, it is the system.
The economic frame matters. In a low-fire, low-hire market, listings are scarce, ghost jobs and scams are common enough that some states are drafting laws against them, and trust between applicants and employers is thin. AI tools on both sides are being deployed to compensate for that thinness, and they are hardening it instead. Employers see a wall of AI-written cover letters and reach for AI ranking. Candidates see a black box and reach for keyword tools. Neither side is measuring whether any of it improves hires.
The market implication for the AI industry is worth watching. Résumé-optimization SaaS is a growing category built on an assumption — universal ATS AI ranking — that Doist's internal test and Toshiba's human-first workflow both undercut. If Greenhouse's own CEO is publicly telling candidates the answer is not more of what the tools sell, the ceiling on that category may be lower than its subscription pricing suggests. The more durable AI hiring product will be the one that helps humans on either side make a better decision, not the one that helps each side generate more volume for the other to filter.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
