Applications Per Job Doubled. Recruiter Attention Fell 79%. Fill Rates Barely Moved.

In 2022 the average job posting drew 116 applications. In 2025 it drew 244. Over the same four years, the share of jobs that actually closed with a hire went from 64.67% to 69.76%. Volume more than doubled. The outcome moved five points. That is the entire argument, and almost nobody is running it.

The numbers come from The Hire Standard, the benchmark report Greenhouse published in March 2026 covering more than 6,000 companies and over 640 million applications between 2022 and 2025. It is the largest public dataset on what actually happened to hiring pipelines during the AI application surge, and it is worth reading as a natural experiment rather than a trend report.

Here is the four-year series, unedited:

• Applications per job: 116 → 189 → 223 → 244. Up 111%.
• Recruiters per organisation: 10.43 → 7.23 → 5.44 → 4.62. Down 56%.
• Annual applications per recruiter: 146 → 297 → 522 → 746. Up 412%.
• Jobs closed with a hire: 64.67% → 67.60% → 69.86% → 69.76%. Up 8% relative — about five percentage points.
• Time to fill: 43.64 → 47.12 → 52.97 → 59.67 days. Up 37%.

Read the first three lines and the fourth stops making sense. The pipeline more than doubled. The people paid to evaluate it were cut by more than half. And the share of searches that ended in a hire got better.

The attention calculation

Put a number on how much human evaluation each application now receives. Divide recruiters per organisation by applications per job:

2022: 10.43 ÷ 116 = 0.0899.
2025: 4.62 ÷ 244 = 0.0189.

The ratio between them is 0.211. Attention per application fell roughly 79%.

That figure combines two separate Greenhouse series and holds two things constant that are not perfectly constant — the number of open jobs per organisation, and the hours a recruiter has available. It is directional, not precise, and it should be read as an order of magnitude rather than a decimal. But the order of magnitude is the point. Four-fifths of the human attention that used to land on an application no longer lands on it.

Now run the counterfactual. If the screening pass were the binding constraint on whether a role gets filled — if the depth of the résumé review were doing the work — then removing four-fifths of it should have collapsed the fill rate. Instead the fill rate improved.

There is only one honest reading. The screening step was never where the decision was made. It was where the paperwork was made.

Volume became free, which is exactly what killed it as a signal

The mechanism behind the surge is not mysterious. Greenhouse reported in June 2026 that applications are up 129% since 2023 while open roles have stayed roughly flat, and described the consequence in a sentence worth quoting exactly: "Resumes are polished into sameness, interview answers are rehearsed, and pipelines are crowded with candidates who look right on paper but aren't in reality."

Fortune reported in July that Greenhouse CEO Daniel Chait calls the resulting cycle a "doom loop": candidates spend roughly $20 on tools that fire applications at hundreds of postings, employers respond with AI filters, rejected candidates respond with more automated applications. Both sides are now automating the same low-information exchange at increasing cost.

When a signal becomes free to produce, it stops being a signal. It becomes a cost.

This is the part the hiring market has not absorbed. For thirty years, the implicit theory of recruiting was that a bigger funnel produces a better hire — that if you could just see more people, you would find the right one. The theory was never tested, because generating volume was expensive enough that nobody could push it to the limit. In 2025 somebody did. Applications per job doubled, and the outcome measures moved five points and got 16 days slower. We now have the answer, and it is no.

What this means for a leadership seat

Everything above describes the mid-funnel market. At the executive level the effect is worse, not better, for a reason that has nothing to do with AI.

The economics of contingent and high-volume search import the volume playbook into a seat where volume was never the constraint. A firm carrying 100 open requisitions and filling eight of them is not running a bad process. It is running a process built for a world where more looking equals more finding — and the four-year record now says that world does not exist. We have written before about what that produces on the receiving end: a long candidate list is a sign nobody read the brief, and 291 applications still do not contain the right person.

The scarce input is not candidates. It is specification. Who does this hire actually report into, and how does that person behave in a disagreement? Is this company in transformation mode, or multiplying something that already works? What does the seat require at month 24, when the company is a third larger than it is today? None of those questions are answerable by looking at more résumés, and all of them determine whether the hire is still there in two years — which is where 89% of hiring failures actually originate.

That work does not scale. It cannot be done across a hundred companies at once, which is precisely why the industry does not do it and precisely why it is worth paying for. Ten searches understood deeply, eight to ten filled. Not because depth is a more pleasant way to work, but because the volume lever has now been pulled to its limit in public, and the data says it does not move the outcome.

The 16 days

One number in the series is a straight cost, not an argument: time to fill rose from 43.64 days to 59.67. Sixteen additional days of an empty seat, per role, industry-wide.

For a functional leadership seat, we have priced what an open leadership role costs per day, and for a revenue seat specifically, the CRO vacancy math. Sixteen days is not a rounding error on either. It is the tax the market is paying to process applications that the same dataset says are not deciding anything.

The honest version of the last four years is this. Recruiters absorbed a 412% increase in applications with a team cut by more than half and still improved the fill rate — which is a genuine operational achievement and also an indictment of the metric. If output holds while the input quadruples and the staffing halves, the input was not what produced the output.

So stop counting applications per job. Start counting how many decisions in your last search were made on evidence nobody else in the market had. On most searches the answer is zero, and that is the number that predicts whether the hire is still in the seat in 2028.