AI is not only replacing jobs. It is also raising the cost of finding one.
The strange part is that this is happening while both sides of the hiring market become more efficient. Greenhouse describes candidates using AI to revise résumés in bulk, write cover letters, and prepare for interviews. On the other side, hiring managers say AI can help them screen applications and make faster, better decisions with fewer recruiter resources.
Every participant saves time. The market as a whole becomes harder to use.
Greenhouse captured this contradiction in its 2025 AI in Hiring report. The survey covered 4,136 job seekers, recruiters, and hiring managers across the United States, the United Kingdom, Ireland, and Germany. In the US, 70% of hiring managers said AI could help them make faster and better decisions with fewer recruiter resources. Only 8% of job seekers believed AI made hiring fairer.
The difference is easy to understand. Employers see hours saved. Candidates experience a system that can reject them without ever explaining what it saw.
When every signal becomes suspect
Software engineer Jacky Alciné recently described the human side of that system in “Being kicked out of the tech industry”. He has roughly a decade of professional experience. He has worked on software used by millions of people, designed architecture, and handled production incidents. This is not someone trying to enter the industry for the first time.
Every day brought at least five rejections, and some days at least six versions of “you’re great, but not a fit.” In the article, he put it bluntly: what you submit no longer matters, because other applicants are cheating during the application process.
Greenhouse found the same arms race at a larger scale. Among 1,200 US job seekers, 41% said they had used prompt injection—hidden instructions intended to influence an AI filter—and 52% of those who had not used it were considering it. On the other side, 91% of US recruiters said they had encountered candidate deception, while 34% spent as much as half their week filtering spam and junk applications.
Candidates use AI to multiply applications. Employers use AI to block them. Greenhouse calls it the “AI doom loop.”
China is already part of the same transition. 51job now advertises AI-powered matching and large-model-driven recommendations. Zhaopin offers AI résumé rewriting and AI interview products. BOSS Zhipin says its hiring-specific Nanbeige model powers search, recommendations, AI-assisted communication, and interviews. Anyone who has used 51job (前程无忧), BOSS直聘, or 智联招聘 will recognize the direction: both sides are being encouraged to move faster, even as it becomes harder to tell which signals are real.
That is the central failure. Individual efficiency is rising while market efficiency is falling.
Jacky’s real experience now has to stand in the same queue as mass-generated résumés, polished project descriptions, and AI interview scripts. Once employers assume that any signal may be synthetic, even real experience becomes harder to verify.
Trust moves back to people
Companies will respond by adding friction elsewhere: identity checks, more in-person interviews, and a greater reliance on internal referrals. Greenhouse found that 39% of US hiring managers were already conducting more in-person interviews to verify authenticity.
These methods are slower and more inconvenient, but they add another layer of verification. They give the company one more chance to confirm that the person exists and the experience is real.
The result is that job hunting increasingly becomes what Jacky called a referral game. Trust moves back into personal relationships because the open application channel has become too noisy. For someone without an existing network, the entrance to the interview process gets narrower.
AI was supposed to remove friction from hiring. It did remove some of the old friction: filling forms and screening piles of résumés. But it replaced that friction with mutual suspicion.
Frontier AI labs keep insisting that AI will not create mass unemployment. But job replacement is only the first layer of the employment question. AI is already making it more expensive—in time, trust, verification, and social access—to connect a real person with a real job.
It may take years to decide whether AI has been good or bad for work overall. In employment today, though, I see more negative effects than positive ones.