The AI Resume Arms Race Is Hurting Everyone, Including the Candidates Winning It

Something has broken in the hiring process over the past 18 months, and it has broken on both sides simultaneously. Job seekers have discovered that AI tools can generate tailored, polished applications at scale. Hiring managers have responded with AI screening tools that filter those same applications before a human reads them. Both sides believe they are being efficient. Both sides are making things worse.

Understanding why requires looking at the incentives honestly, without judging the people caught inside them.

Why Candidates Are Doing This, and Why It Makes Sense

The numbers behind the current tech hiring market are genuinely alarming for job seekers. According to LinkedIn data published in early 2026, a single software engineering role at a mid-size company now receives an average of 385 applications, up from roughly 140 in 2022. For roles at large enterprises and GCCs, that number regularly exceeds 800. A job seeker applying to 20 roles the traditional way, each application carefully written, would spend approximately 30 to 40 hours on cover letters and tailored resumes alone, and statistically receive responses to fewer than three of them.

Into this environment came ChatGPT, Claude, and a generation of resume optimisers that promised to level the field. Job seekers started using AI to rewrite their resumes to match specific JD language, generate cover letters in seconds, and apply to hundreds of roles without the manual effort that had previously acted as a natural limit.

The Signal Collapse

The problem is not that candidates are lying. Most of them are not. The problem is that the signal has collapsed. When every application is polished to the same standard, calibrated to the same keywords, and formatted for the same ATS criteria, the differentiation that used to exist between candidates disappears. A hiring manager looking at 400 AI-assisted applications sees 400 applications that look roughly the same. The resume that once told you something about a candidate's judgment, priorities, and communication style now tells you almost nothing.

Research from Harvard Business School published in late 2025 found that recruiting managers rated AI-generated resumes as higher quality than human-written ones in blind comparisons, but reported greater difficulty distinguishing between candidates when all applications had been AI-enhanced. The quality floor went up. The signal ceiling came down.

Hiring managers responded in the predictable way: they added more AI screening layers. ATS systems now reject an estimated 75 percent of applications before any human reads them, according to a 2025 report from Harvard Business School's managing the future of work project. Those systems filter for keyword density, formatting compliance, and increasingly, AI-scoring of cover letters. The result is an AI screening system sorting through AI-generated applications to find candidates who may have had their genuine skills buried under the optimisation layer.

What This Actually Costs Candidates

Here is the part that gets missed in the conversation about AI resumes: the candidates who "win" this arms race are not winning much.

A candidate who games the screening process and lands an interview for a role that was not actually a strong fit has only delayed the discovery of that mismatch. They will either fail in the interview when domain-specific questions reveal the gap, or worse, they will pass the interview, accept the offer, and discover 60 days in that the role is not what they needed it to be. That is a resignation, a gap in their work history, and a hiring process that starts again from scratch for a role that should never have been filled that way.

More subtly, the mass-application approach trains candidates to think of job searching as a volume problem rather than a targeting problem. That mental model is wrong. Tech hiring, particularly at the senior level, is not a lottery. The candidates who get the best outcomes consistently are the ones who identify roles where their specific background is genuinely differentiated, and then pursue those roles with real intent.

What Still Works, and Why

Genuine fit still gets hired. The evidence for this is straightforward: companies are still making tech hires, senior roles in particular are still being filled, and the hiring managers doing the filling are not selecting randomly from an undifferentiated pool. They are selecting the candidates who, through some combination of referral, research, or clear signal, demonstrate that they understand the role specifically and are the right fit for it.

The candidates who do this well share a few characteristics. They apply to fewer roles, but those roles are more targeted. They research the company, the team, and the specific challenge the role is addressing before they write anything. Their cover letter, whether AI-assisted or not, is specific enough that it could not have been sent to a different company. And they are honest about what they know and what they are still developing, because that honesty is actually more compelling to an experienced hiring manager than a resume that claims mastery of everything.

"The candidate who tells you precisely what they are good at and precisely where they want to grow is far easier to hire confidently than the candidate whose resume says they can do everything."

What Employers Need to Change

The arms race is not only a candidate problem. Employers created the conditions that make mass AI application rational by building screening systems that filter for keywords rather than judgment, by posting JDs that list 18 requirements for roles that actually need six, and by making the application process so opaque that candidates reasonably conclude that volume is their only lever.

The structural fix is not better AI screening on the employer side. That just escalates the arms race further. The fix is bringing human judgment further forward in the process. That means a meaningful intake conversation before sourcing begins, so that the sourcing targets the right people rather than accepting whoever opts in. It means shortlists built on genuine assessment rather than keyword matching. And it means briefing candidates honestly about the real role before they decide whether to invest time in the process.

At Qfyre, the intake session happens before we source a single profile. We understand what the role actually requires, which means we can recognise it when we see it, rather than filtering for signals that correlate with the requirement at best. The result is a shorter funnel with a higher conversion rate, which is better for everyone including the candidates who are not right for this role but would be right for a different one.

The arms race will not be won by either side. It will be ended by hiring systems that make volume competition unnecessary because the matching is precise enough that candidates know whether they are right for something before they apply. That is a solvable problem. But it requires employers to invest in the intake end of the process, not just the screening end.