Greenhouse processes over 60 million applications a quarter. Its CEO, Daniel Chait, recently put a number on what that volume has done to the recruiting funnel: applications per recruiter are up 412 percent since 2022, and the average job posting on the platform now draws around 254 applicants. Chait has a name for what happens next. He calls it the AI doom loop.
The mechanism is not complicated. A job seeker pays roughly $20 for an auto-apply tool. It fires off hundreds of applications a week with no real intent behind any single one. Recruiters get buried, turn on AI screening to survive the flood, and the filters get tighter. Candidates who keep getting filtered out respond the only way the system has taught them to: they apply to more jobs. The loop closes on itself, and as Chait told Fortune, it is the first time both sides of the hiring relationship have been miserable at the same time.
That is the headline. The part that matters more for anyone running a hiring function is what the volume actually did to the funnel underneath it. Applications went up more than 400 percent. Hiring did not go up 400 percent. It did not go up at all in most functions. The volume increase did not produce more qualified candidates reaching decision-makers. It produced more noise for recruiters to filter before they could find the same number of qualified candidates they were already finding.
Most hiring leaders are still measuring recruiter output the way they measured it five years ago: applications received, requisitions opened, time spent screening. Those metrics assume application volume and candidate signal move together. They do not anymore. They have decoupled.
We think the right way to describe what is happening is a Signal Tax. Every unit of application volume above a certain threshold costs a recruiter real time and produces a shrinking marginal return in actual hireable candidates. Below that threshold, more applications mean more options. Above it, more applications mean more filtering, more false negatives, and a slower path to the person who was actually right for the role. Greenhouse’s own webinar data with Scede shows average applications per role climbing from roughly 150 to 746. Nobody is claiming the quality of the candidate pool behind that number climbed five times over. It didn’t. The tax got paid in recruiter hours and in qualified candidates who got lost in a haystack that used to be a stack.
This is not a sourcing problem. Sourcing was never the constraint. The constraint was always finding signal fast enough to act on it before a competitor does. AI-generated volume attacks exactly that constraint, on both sides of the table at once.
Three things follow from this, and each one changes an actual decision an executive is making right now.
First, application count is no longer a proxy for pipeline health, and treating it as one will get you the wrong read on your team’s performance. A recruiter triaging 746 applications for one opening is not doing more work than a recruiter who received 150 well-targeted ones. She is doing worse work, because the ratio of signal to noise she is filtering through has collapsed. If your team’s dashboards still lead with applications received, you are rewarding the wrong behavior and possibly burning out the people best at spotting real signal, because they are the ones drowning fastest.
Second, screening automation bought to manage the flood is not neutral. It is a second-order cause of the same collapse. AI screening tools built to survive volume tend to optimize for pattern-matching against keywords and formats, which is exactly what AI-written applications are optimized to produce. The two systems are now training against each other, and the candidates most likely to get filtered are often the ones who did not game the format, not the ones who are least qualified. If your AI screening layer was adopted purely as a defensive measure against volume, it is worth an honest audit of who it is actually surfacing versus who it is quietly removing.
Third, the fix is not more filtering. It is a different starting point. The organizations that are not stuck in the loop are largely the ones that never entered the high-volume, inbound-application funnel to begin with. They are sourcing directly, building relationships with a small number of qualified people before a role opens, and converting a much higher share of first conversations into hires. That is a structurally different model from posting a role and managing what shows up. It requires giving up the illusion that a bigger applicant pool is a safer applicant pool.
The instinct across the market has been to fight volume with more automation on the receiving end. That instinct is understandable and it is also the second half of the loop, not the exit from it. AI that scales submission does nothing for match quality on either side of the transaction, and stacking AI screening on top of AI-generated applications does not restore signal. It just moves the noise one step downstream and calls it progress.
The question worth asking internally is not how to process 746 applications faster. It is whether your team should be receiving 746 applications for that role in the first place, and if the honest answer is that inbound volume has become the default hiring channel by accident rather than by design, that is the actual finding.
We do not run inbound funnels, so we see this from the other side of the table. Our average time to fill sits around 33 days, against a SHRM benchmark closer to 44, and our submittal-to-hire ratio runs about 8:1 in senior technical search, against an industry norm of 10:1 or worse. Those numbers exist because the process starts with direct outreach to a small number of people who match the role, not with a job posting absorbing whatever volume shows up. We are not immune to AI-generated noise in candidate communications, but it never enters our pipeline in the first place, because the pipeline is built candidate by candidate rather than filtered from a pool of thousands.
The pattern we are seeing in client conversations right now is a shift in the question being asked. Six months ago, hiring leaders were asking how to screen faster. Increasingly, they are asking whether the inbound channel is worth running at all for senior and technical roles, given what it now costs to filter. That is the right question. It just was not the question most teams were set up to answer with the metrics they had in place.
If your team is measuring recruiter performance by volume processed rather than signal converted, that is worth a direct conversation before your next senior search opens.
Verticalmove is a strategic talent consulting partner that helps organizations solve business problems related to talent attraction, selection, and retention. We work with PE-backed, venture-backed, mid-market, and enterprise companies to design talent strategies, strengthen leadership teams, and build the workforce capabilities required to achieve critical business objectives. When growth stalls, transformation accelerates, or organizational priorities shift, talent is often the constraint. We help companies identify, attract, assess, and retain the people who create competitive advantage.