AI hiring doom loop: the shortlist missed the people they hired

A remote company called Doist ran an experiment worth more than the phrase everyone is quoting. It took roles it had already filled, fed the job descriptions and the saved applications into AI shortlisting, and checked whether the people it actually hired came out on top. In both tests they did not. The hire was missing from the short list.

What everyone is optimising for

That result sits underneath the whole business of optimising a CV for software. The belief driving it is specific: a hiring team runs applications through an applicant tracking system, an AI ranks them, and only the top 10 to 20 percent get read by a person. Tools that promise to beat the ranking charge $30 to $50 a month for keyword tailoring and formatting advice.

The belief is not always right. Some companies rank automatically. Others keep hiring in human hands from the first application to the offer. One recruiter in the Wired report scored terribly on the optimisation tools and still collected a dozen interviews and an offer.

“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, Wired, 4 September 2026

Quote source: Wired, 4 September 2026

The split is cultural, not technical

Employers are stuck in the same trap from the other side. They open a posting in a market with few openings, receive hundreds of applications that read almost identically because candidates polished them with the same models, and reach for automated ranking to tell them apart. The output of one side's AI becomes the input of the other's.

The split between the two camps is cultural rather than technical. The reporter spoke with dozens of recruiters, HR managers and owners who hire for themselves, and the divide tracked neither company size nor the number of applications received. Kim Jones, vice president of human resources at Toshiba, said her team has humans review every application, and that polishing with AI will not help anyone get past the system.

Where the AI hiring doom loop costs you

Where AI does cost a candidate is later. Jones described what an assisted interview sounds like from the other end of the call: the pause, the typing, and then a long answer that arrives too neatly. Nothing in that sequence needs a detection tool.

Two things follow for anyone job hunting inside this loop. A subscription that optimises for a filter you cannot confirm exists is a bet, and the company selling it collects while you keep searching. And the interview stays the part where a person is still doing the judging, which is the opposite of where most of the effort currently goes.

The wider pattern is the same one we keep writing about. Corporations deploy agents faster than they can measure them, and staff push back where they can, as thousands did in the Sydney university strike over job security. Hiring is the first place where both sides got the same tool at the same time.

Nothing here should be taken as financial advice; treat it as information to consider.