Make talent quality your leading analytic with skills-based hiring solution.
The AI hiring arms race describes a specific, self reinforcing dynamic: candidates use AI tools to apply to far more jobs than they ever could manually, employers respond by using AI to screen and filter that flood, and neither side ends up better off. Greenhouse CEO Daniel Chait has described the result bluntly as an “AI doom loop,” noting that everyone is using their own AI to solve their own problem, but it is making the whole system worse for both sides at once. The numbers back that up.
Across roughly 175,000 live positions on Greenhouse’s platform, the average job now receives around 254 applications, a 412 percent increase in applications per recruiter. In one striking example, 1.2 million applications were submitted in the UK for fewer than 17,000 graduate roles in a single cycle. This is what the AI hiring arms race actually looks like in practice, not a metaphor, a measurable flood on both sides of the table.
Job seekers have adopted tools that automatically apply to every matching posting on a platform, often for a flat fee, submitting applications at a volume no human could sustain manually. One widely cited case involved a job seeker with a master’s degree who applied to thousands of positions over six months without a single callback, illustrating that raw volume does not translate into results on its own.
Ironically, candidates using structured, verified application features built by the hiring platforms themselves, rather than blind mass apply tools, have reported callback rates around five times higher than everyone else, a signal that verified signal beats raw volume even inside this dynamic.
On the other side, recruiters facing hundreds or thousands of applications in a single day or two increasingly rely on AI screening tools to make the volume manageable at all. The problem is that AI generated applications, often built from the same templates and the same optimized keyword patterns, increasingly all start to look the same to both the AI doing the screening and the human reviewing the shortlist.
This mirrors the same signal collapse problem covered in more depth in our piece on skillfishing, where a strong looking application on paper increasingly tells a recruiter very little about actual ability.
This is the core mechanic of an arms race: each side’s AI adoption is a rational individual response to the other side’s AI adoption, but the combined effect degrades the signal everyone actually needs.
Chait has pointed out this is the first time in recent memory that both job seekers and recruiters have been simultaneously unhappy with the hiring process at the same time, which is a useful diagnostic, a system where both sides are losing is not a system that a little more of the same technology, applied by either side, is likely to fix.
Breaking the loop requires shifting the deciding signal away from application volume and resume keyword matching entirely, toward something an AI generated application cannot fake: a verified, structured demonstration of actual skill.
This is the same underlying argument made in our companion piece on AI interviewers, and it is the direct rationale behind pairing any AI screening step with a real skill assessment platform, so the deciding factor becomes demonstrated ability rather than who submitted the most applications the fastest.
The AI hiring arms race is not a temporary glitch that will resolve itself as tools mature, it is the predictable outcome of both sides individually optimizing for volume in a system that only has room for one winner at a time. The way out is not less AI, it is redirecting what AI on the employer side actually measures, from resume volume and keyword density to verified, demonstrated skill.
The AI hiring arms race describes candidates using AI tools to mass apply to jobs while employers use AI to mass screen the resulting flood of applications, a dynamic that Greenhouse’s CEO has called an AI doom loop because it leaves both sides worse off.
AI powered auto apply tools let candidates submit far more applications than manual applying ever allowed, pushing average applications per job into the hundreds and driving a 412 percent increase in applications per recruiter on some platforms.
Many candidates use tools that automatically submit applications to every matching posting on a platform, sometimes for a flat fee, though evidence suggests verified, platform native application features produce meaningfully higher callback rates than blind mass apply tools.
Recruiters increasingly rely on AI screening tools to filter large volumes of applications, but AI generated applications built from similar templates increasingly look alike to both the screening AI and the human reviewer, reducing the tool’s usefulness.
Employers can shift the deciding signal away from application volume and resume keyword matching toward verified, structured skill assessment, since a demonstrated skill result is much harder for an AI generated application to fake than a keyword optimized resume.
Synthetic identity fraud in hiring happens when someone builds a job candidate instead of being one, blending a real Social Security number or stolen document with a fabricated name, a generative AI face, and a scripted background, so the “person” who interviews and gets hired never actually existed as a single, real individual. It is […]
A diploma mill is an unaccredited operation that sells degrees or certificates for a flat fee, little or no coursework, and almost no academic oversight, and the fastest way to catch one is to verify education claims directly with the issuing school rather than trusting the document a candidate submits. Fake degrees are no longer […]
Nation state hiring fraud is when a government backed operative uses a stolen or fabricated identity to get hired into a normal remote job, most often in IT, so that their wages, and sometimes the access that comes with the role, can be funneled back to a sanctioned regime. The best documented version of this […]
