The number that defines hiring in 2026 is a rate: roughly 11,000 job applications per minute, submitted through LinkedIn alone, according to the platform’s own figures reported by The New York Times. That volume is up around 45 percent in a year. Over a comparable stretch, LinkedIn’s data showed job postings falling by more than ten percent.
Hold those two curves in mind together. Applications surging while openings shrink means the ratio of applicants to jobs is not drifting upward, it is being inverted, and the extra volume is not coming from a sudden global increase in effort. It is coming from the collapse in the cost of applying. Generative tools now produce a tailored CV in seconds, a bespoke cover letter in one prompt, and answers to screening questionnaires without the candidate reading the questions. Auto-apply services go further, spraying applications across hundreds of postings while the applicant sleeps. Reports describe single job postings drawing over a thousand AI-generated applications within days.
Recruiters have a name for what arrives: hiring slop. This piece is about what the flood is doing to the hiring funnel, why the standard responses are making it worse, and why the discipline this newsletter cares about, verification, is about to change its position in the process.
How the Flood Rose
Blame the incentives before the people, because from the applicant’s side, every step of this was rational.
The labour market context supplied the desperation: US employers cut more than 1.17 million jobs in 2025, the most since the pandemic, pushing waves of experienced people into a shrinking opening count. The tools supplied the means: Gartner’s surveying found roughly four in ten candidates already acknowledge using AI in the application process, mostly to generate CV text, cover letters, and assessment answers, and that is the admitted figure. And the arithmetic supplied the logic: if one application costs three minutes instead of three hours, sending three hundred is not madness, it is optimisation. Workday’s platform data caught the same wave early, with applications up 31 percent in a half-year against job growth of seven percent.
The applicant tools work exactly as advertised. That is the problem. Their success is measured in applications sent, and the cost of every application sent is borne by someone else: the recruiter who must process it, and the genuine candidate buried beneath it.
The Collapse of Signal
To see what the flood destroys, remember what an application used to be.
A tailored CV and a considered cover letter were costly signals. The cost was the information: only someone genuinely interested in this role, with genuine relevant experience to organise, would spend the evening producing them. Recruiters never really read applications for their prose. They read them for the effort the prose implied.
Generative tools removed the cost, and the signal died with it. Recruiters now describe stacks of near-identical résumés, each instantly optimised to the posting’s keywords, indistinguishable in polish, and useless for telling genuine interest from spray. One founder put the practitioner’s despair plainly: it is impossible to know who is actually qualified, and increasingly impossible to know who is real. Some recruiters claim they can spot AI-written applications in under twenty seconds, and perhaps they can, but the claim misses the point. At nine hundred applications per posting, twenty seconds of detection each is five unpaid working days, and the well-prompted fakes are not the ones getting caught.
Meanwhile the keyword arms race grinds on. With applicant tracking systems on effectively every Fortune 500 career site, candidates optimise their text for the machine, the machine filters harder, and the optimisation tools adapt again. Both sides are getting better at a game that produces no information.
Bot Versus Bot
The employer response, so far, has mostly been symmetrical escalation: fight the flood of machine-written applications with machine-run screening.
The momentum is real. LinkedIn’s research finds 93 percent of recruiters planning to increase their AI use this year. The platform itself ships a Hiring Assistant that automates job descriptions, candidate screening, and messaging. Chipotle credits its AI recruiting system with cutting hiring time by 75 percent. And to be fair, the tools have genuine wins: 60 percent of recruiters say AI has surfaced hidden-gem candidates their manual search would have missed, which is what machines are actually good at, retrieval from a haystack.
But as a complete answer, automation-versus-automation is a dead end, and the industry’s own commentators have named it: an AI-versus-AI standoff, spiralling. Screening models are unreliable in exactly the ways that matter, and every improvement in filtering trains the application tools that feed it. Worse, the crossfire is landing on candidates. Gartner finds only 26 percent of applicants trust AI to evaluate them fairly, a quarter say employer AI use makes them trust the employer less, and offer acceptance has collapsed from 74 percent to 51 percent in two years. Half of candidates are not even sure the jobs they apply to are real, and they have reason: the FTC tracked job-scam losses rising from 90 million dollars in 2020 past 501 million in 2024. The flood runs in both directions, fake applications meeting fake postings, and the residue is a market where each side assumes the other is machinery.
Readers of the recent piece on compensation will recognise the pattern: an equilibrium of mutual distrust, where both sides discount everything and honesty feels like a handicap. Hiring’s front door has caught the same disease as its salary negotiation.
Where the Fraud Slips In
Now the part that concerns this newsletter directly. A funnel drowning in noise is not just inefficient. It is cover.
When no human meaningfully examines the top of the funnel, nobody verifies anything there either, and the population that benefits most is the one with something to hide. At the mild end, that means inflated claims travelling further than they used to, because the AI-polished fabrication looks exactly like the AI-polished truth. At the sharp end, it means applications not attached to real candidates at all, the synthetic fringe this newsletter has covered before, for whom a flooded funnel is ideal habitat. Gartner’s warning bears repeating: it is getting harder to evaluate candidates’ true abilities, and in some cases their identities, and the firm projects one in four candidate profiles will be fake by 2028.
The flood also feeds directly into the assessment-integrity crisis covered in this series’ recent proxy-interview work. When 20,000 candidates sit a single company’s assessments, as India’s campus pipelines routinely produce, volume is precisely what impersonation and malpractice hide inside. Noise at the top of the funnel becomes fraud in the middle of it.
The Return of Costly Signals
Markets flooded with cheap talk always respond the same way: they reprice the signals that still cost something.
It is already visible. Reporting in the Wall Street Journal describes companies including Cisco and Google reinstating in-person interviews, partly to confirm that candidates are who they claim. Talent leaders describe a broader flight to quality: pulling back from fully automated pipelines toward processes where verified humans meet verified humans. Referrals are regaining weight. Professional networks now offer identity verification badges, an acknowledgement from the platforms themselves that “profile” and “person” have come apart. Work samples and live assessments are displacing CV prose, provided the assessment itself is protected by the integrity controls this series has already described.
Notice what unites these responses. None of them out-generates the generators. They step outside the text war entirely, to things that cannot be prompted into existence: presence, identity, and verified fact. That is the asymmetry that decides this arms race. Generation scales infinitely for the applicant. Verification scales for the employer.
Make the Funnel Cost Something True
Which brings the argument home to screening, because the flood quietly breaks the assumption traditional background verification was built on.
BGV’s classic position is post-offer: one finalist, checked thoroughly, after the funnel has done its filtering. That design assumed the funnel upstream was mostly honest and mostly human, so the expensive check could wait for the end. Neither assumption survives 11,000 applications a minute. The costly mistakes now happen earlier: interview loops spent on candidates whose materials were written by something that never met them, assessment slots consumed by impersonation, shortlists contaminated before any check has run.
The response is not to run full screening on nine hundred applicants, which is impossible, but to move small, cheap verification forward: identity anchored before anyone performs, key claims spot-checked at the shortlist, verified candidates fast-tracked. Instant statutory-record checks have made this economically trivial in ways that were not true five years ago. The full check still guards the offer. The funnel, meanwhile, gets what it lost when applying became free: a gate that costs nothing to the honest and everything to the fabricated, a point where the words stop and the facts begin.
The companion playbook builds that architecture step by step. The principle fits on one line: when talk becomes free, verify.
