AI is reading AI and hiring is stuck

There is no live venue pricing this hiring doom loop, no current odds and no traded volume to watch, and the only thing that moved in the story is the count of applications from 38 per graduate vacancy 20 years ago to 140 last year. That leaves the debate priced on surveys and anecdote, not a liquid market, which limits what you can infer about adoption. Not financial advice. Markets move fast.

BBC News opens on the loop everyone feels but few can measure, where candidates use AI to write more applications and employers use AI to sift them.

It feels like talking to a robot.

The piece follows job hunters Neve, Addie and Santina, who describe sending 50, 100 and more than 200 applications in the last year, and it pairs them with Kate Young, who runs an AI consultancy, and Matt Weston, a senior executive at recruitment firm Robert Half, which places over 100,000 candidates a year globally. One survey in the report puts AI use at 73% of students and graduates at some stage of the process, mostly to tune CVs and cover letters for keywords that an automated sift might hunt for, and the news peg is simple, vacancies are down, graduates are at a record high and polish is now free.

Here is the gap the skeptics will notice. 75% of candidates believe AI is screening CVs, but only 21% of recruiters reported using it to do so, a spread that suggests perception is ahead of deployment. BBC News also notes the Information Commissioner says these tools can help process high volumes consistently and quickly, while Young argues human hiring has always been flawed and that AI could screen out noise and bias if built well. Then comes the caveat she stresses herself, that the guardrails are many and complicated, from who built the system to which technique was used, and that is a big if that the piece leaves unproven.

The same unnamed sameness shows up on the employer side. Weston says AI can generate a CV from a job spec, and applications now look very similar, so the work shifts to verifying whether the paper matches the person. Researchers call the wider risk algorithmic monoculture, where different employers using the same or similar models may favor the same candidates again and again, a concern Stanford work links to the reuse of a single vendor score across many employers. The BBC News story flags the referral workaround that follows, when polish tells you nothing, personal recommendations regain weight, and that favors networks over novelty.

The interview stage makes the tension concrete. If you pass the sift, you may meet an AI interviewer, sometimes an avatar that replies to your answers, sometimes a set of timed prompts to camera. The World Economic Forum is cited as saying AI interviews can expand opportunities and reduce bias, but the candidates in the piece describe the inverse, one was told by an AI it lacked empathy, others call the process uncanny and inhumane and say they never feel understood. You want to get in front of a human before you even get to compete for the job, and that connection is now gated behind automation that candidates feel judges keywords, not intent.

For startups selling the plumbing, the pitch is scale, not certainty. HireVue describes itself as an AI-driven suite for structured video interviewing and skills testing and assessments for every level and role, and says 1,150 companies use it, with marketing claims of a 90% decrease in time to hire. That context matters because BBC News does not name the vendors behind the sifts or interviews it discusses, nor does it show validation data, error rates or how an employer would tailor a model to a specific role without importing someone else's bias. Until those details are disclosed and auditable, candidates will keep playing the keyword game, employers will keep vetting tightly, and AI will keep reading AI.