The volume problem

The résumé worked because it signaled effort. Crafting one took time. Tailoring it to a specific role showed genuine interest. The document was proof of work.

That premise is dead.

ChatGPT can generate a keyword-optimized résumé from a single job description paste in under 30 seconds. AI agents, services that autonomously find and apply to jobs for you, can submit hundreds of these in a day. The result is what Ars Technica accurately called “hiring slop”: a flood of applications that are grammatically flawless, structurally identical, and almost entirely devoid of signal.

Over 90 percent of job seekers now use generative AI in their applications, per Huntr’s Job Search Trends Report. Meanwhile, 91 percent of US employers deploy AI somewhere in their hiring workflow. Both sides optimized for throughput, not judgment. The gate is wide open, and everything gets through, which means nothing does.

The numbers confirm it. Kaiser and Dabulis independently estimate that among hundreds of applicants for a typical role, fewer than five are well-suited. On LinkedIn, 70 percent of hirers say less than half the applications they receive meet all the criteria. The system produces volume at the expense of signal, and both sides pay the cost in wasted time.

The arms race

Companies didn’t sit still. Their response was predictable: fight AI with AI.

Chipotle’s chatbot screener, Ava Cado, reportedly cut hiring time by 75 percent. Resume Now reports that 99 percent of Fortune 500 companies use applicant tracking systems that reject roughly 75 percent of résumés before any human sees them. The candidate experience becomes: AI writes the application, AI screens the application, AI rejects the application. No human was ever in the loop.

And it escalates. Candidates use AI to generate interview answers. Companies use AI to detect AI-generated answers. Gartner estimates that by 2028, roughly one in four applicants could be fraudulent, a projection that includes identity theft rings, North Korean IT workers using stolen credentials, and deepfake-enabled video interviews. The Justice Department indicted two North Korean nationals in January for exactly this kind of scheme.

Anthropic, of all companies, recently advised job seekers not to use LLMs on their applications. Striking admission from a company whose entire business model depends on people doing the opposite. Even the people selling the tool recognize that when everyone uses it, nobody benefits.

This is a machine-to-machine negotiation where the humans pay the tab. Candidates use AI to generate text optimized for the ATS. The ATS uses AI to rank candidates by keyword density. The candidate’s AI adds more keywords. The ATS adjusts its weights. A recursive loop with no information value. The hiring equivalent of two modems handshaking into eternity.

What survives

If the text-based résumé has degraded into noise, three things still carry signal.

First: referrals. Kaiser is direct. “Referrals are going to be the way of the future.” When everyone can generate a perfect application, the only differentiator is someone inside the building willing to vouch for you. A referral bypasses the ATS entirely, substituting a machine judgment for a human one. In an environment where 75 percent of résumés are rejected before they reach a person, getting placed directly in front of that person is the difference between being seen and being noise.

Second: demonstrated work. A portfolio, a GitHub repo, a case study, a published piece, a project with measurable results. These are things AI cannot hallucinate into existence for you. It can generate the output, but it cannot generate the history of having done the work. Hiring managers tired of reading 400 identical résumés are increasingly looking for artifacts of real output. The Polsky Center at the University of Chicago recommends structured short-form video responses and skills-based assessments early in the hiring funnel. The question shifts from “Can you write a résumé that matches this job description?” to “Can you show me what you’ve actually built?”

Third: specificity only a human would write. This is subtle but powerful. AI-generated applications are generic at the level of detail because they train on what is common, not what is specific. A real application includes the thing that does not fit neatly into a bullet point. The failed project that taught you something. The niche tool nobody asks for. The unusual career transition. These are negative signals to an AI screener and positive signals to a human reader. If you want to signal “I am a real person,” include the thing an AI optimizer would advise you to cut.

Advice for job seekers

You cannot win the AI optimization game. If you try to beat the machines at generating optimized text, the machines will always win. They are faster, cheaper, and better at it than you. The only strategy that works is to stop playing that game entirely.

What actually works in May 2026:

Stop mass-applying. Easy Apply and AI auto-submit tools are the fastest way to become invisible. Dabulis submitted 700 applications and heard nothing. When you spray your résumé across 200 jobs, you are not increasing your odds. You are distributing your signal so thinly that none of it registers. Every recruiter I spoke to said the same thing: quality over quantity has never mattered more.

Target companies, not job listings. Identify 10 to 15 companies where your skills fit. Spend your time understanding their business, their recent moves, their pain points. Then find a way in: a referral, an introduction, a thoughtful cold message that proves you did the homework. This takes longer per company, but the hit rate is immeasurably higher than spray-and-pray.

Build public work. If your field allows public output, put it where people can see it. Writing, code, design, analysis, strategy. Whatever form it takes. A published case study beats a perfect résumé every time. A GitHub repo with real commit history beats a certificate. A blog post that shows how you think beats a cover letter written by ChatGPT. The bar for “this person clearly exists and can do the thing” is shockingly low right now, which makes it your best investment.

Lead with the signal, not the document. When you submit an application, lead with the thing that cannot be fabricated. Include a link to your portfolio in the first paragraph of your cover letter. Reference a specific problem the company faces and how you would approach it. Make the first thing a human sees, if you are lucky enough for a human to see it, something that could only have come from you.

The bottom line

The hiring signal is not broken forever. It is just broken in its current form. The paper résumé served as a meaningful proxy for effort and interest for decades. AI rendered it useless using the very technology the résumé helped popularize. The market is correcting. Companies are experimenting with work-sample tests, video introductions, skills-based assessments, and paid trial periods. Candidates who adapt to these new signals will have an advantage. Candidates who keep optimizing for the old ones will become increasingly invisible.

The companies that solve this first, that stop trying to screen AI with AI and start designing processes that surface real human capability, will win the talent war. The candidates who solve it first, who stop optimizing for the machine and start showing actual work, will win the jobs.

The résumé has been demoted from the main event to a formality. What matters now is what you can show, not what you can write.

Morgan Chase is Career Strategy Analyst at Duelling Hares. This piece was researched in May 2026 using publicly available data from LinkedIn, the New York Times, Ars Technica, CNBC, Resume Now, Huntr, the Polsky Center at the University of Chicago, and Gartner.