AI Killed the Resume

100 Years of Science Shows What Should Replace It.

Trevor Higgs | March 2026

---

The resume is dead. AI killed it.

Not all at once, but through a slow, escalating arms race that has rendered traditional application screening effectively meaningless. 72% of resumes[1] are now skills/keyword-optimized. Candidates use AI tools to reverse-engineer job postings and generate perfectly tailored applications in minutes. Employers respond with more aggressive AI filters. Candidates counter with more sophisticated optimization tools.

The cycle escalates. Nobody hires better. Everyone works harder.

The Casualties of the Doom Loop

The damage extends far beyond recruiter frustration. 91% of recruiters now report spotting deception [2] in AI-optimized applications, but spotting it and stopping it are different challenges entirely. 34% spend half their work week filtering junk application [3]s, time that should be spent evaluating genuine talent.

The most devastating casualty: early-career hiring has collapsed by 73% [4]. Not 7%, seventy-three percent. Entry-level tech job postings fell 67% between 2023 and 2024. Recent graduates now represent only 7% of new hires, down from 11%.

This isn't because young talent doesn't exist. It's because they don't yet have the keyword-optimized resumes, the years of experience, or the AI-crafted narratives that survive algorithmic screening. The very tools designed to improve hiring efficiency are systematically excluding an entire generation of workers.

The 100-Year Answer

While the industry chases the next AI breakthrough, the most powerful predictor of job performance has been validated for a century.

The Schmidt meta-analysis synthesized over 100 years of industrial-organizational psychology research [5], evaluating 31 different selection methods across hundreds of thousands of employees and job categories. The finding was unequivocal: general mental ability (GMA) is the single most accurate predictor of job performance, with a validity coefficient of .65 out of 1.0.

To put that in context: GMA is approximately 4x more predictive than years of experience (validity: .16) and roughly 2.5x more predictive than reference checks (.26). Combined with structured interviews, the predictive validity reaches r = .76, the highest of any combination ever studied.

This isn't a niche finding or a preliminary result. It's the most extensively validated conclusion in the history of personnel selection research, replicated across industries, job levels, and cultures.

Why Potential Can't Be Gamed

The critical advantage of cognitive potential assessment is its resistance to the gaming that has made resume screening unreliable.

Resumes can be keyword-stuffed. Personality tests can be researched and manipulated, candidates can find the "right" answers for any role with a quick search. Interview skills can be coached and rehearsed. Even video interview AI can be optimized against by candidates who understand the behavioral cues these systems evaluate.

Cognitive potential assessment measures fundamental capabilities, spatial visualization, verbal ability, arithmetic reasoning, perceptual speed, fluid reasoning, that reflect genuine cognitive capacity. You can't keyword-stuff a problem-solving test. You can't AI-generate your way through a pattern recognition assessment.

The Diversity Enable

There's a dimension to potential-based assessment that doesn't get enough attention: it's fundamentally more equitable than resume-based screening.

When you stop filtering for skills/keywords, years of experience, and institutional credentials, you open the door to candidates that traditional screening systematically overlooks. Career changers who bring valuable cognitive capabilities from different industries. Bootcamp graduates who chose alternative education paths. Workers from underrepresented backgrounds who had fewer opportunities to accumulate the "right" resume signals.

86% of employers now express confidence in hiring bootcamp [6] graduates, but keyword-matching systems still filter them out. Potential-based assessment evaluates what these candidates can do, not where they've been. It doesn't care about pedigree. It measures potential.

The Career Quotient Model

This is exactly what Career Quotient (CQ) was built to do. A single, job-specific score from 1 to 100 that predicts a candidate's ability to replicate the success of top performers in a specific role. One person takes a 15-to-20-minute assessment once and receives a CQ score for every role in the company.

CQ models top performers to create data-driven profiles of what success looks like. Then it matches candidates — internal or external — against those profiles based on cognitive potential, career interest alignment, and work values. The result: hiring decisions backed by science, not resumes.

The AI doom loop was inevitable once both sides of the hiring equation had access to the same optimization tools. The exit isn't better filters — it's measuring what can't be gamed

---

[1] HireVue and Aptitude Research. (2025). 2025 Global Hiring Report.

[2] Greenhouse. (November 2025). 2025 AI in Hiring Report. Over nine in ten recruiters have spotted candidate deception.

[3] Greenhouse. (November 2025). 2025 AI in Hiring Report. 34% of recruiters spend up to half their week filtering spam and junk applications.

[4] Ravio. (2025). 2025 Tech Job Market Report. Entry-level positions (P1 and P2 job levels) saw a 73% decrease in hiring rates.

[5] Schmidt, F. L., Oh, I.-S., & Shaffer, J. A. (2016). The Validity and Utility of Selection Methods in Personnel Psychology: Practical and Theoretical Implications of 100 Years of Research Findings. Working paper. (Original meta-analysis: Schmidt, F. L., & Hunter, J. E. (1998). Psychological Bulletin, 124(2), 262-274.)

[6] Course Report and TripleTen. (2024). 2024 Employer Survey on Hiring Bootcamp Graduates.