Job applications written with AI may perform better against AI-driven screening systems than human-drafted ones — an asymmetry Inc. has labeled a quirk every applicant should know. The finding, framed by the outlet as "AI likes to use AI," touches the core of automated hiring: the tools employers deploy to filter candidates appear to reward text produced by tools much like themselves.

For HR leaders, this is not a curiosity. It is a process-integrity question. If algorithmic screeners (software that scores résumés and applications against keyword and formatting models) systematically favor machine-generated submissions, then the ranking a recruiter sees is partly a measure of who used AI — not who can do the job.

What does the quirk mean for hiring quality?

The mechanics are straightforward. AI-generated applications tend to mirror the language patterns, structure and phrasing that screening algorithms were trained to recognize as strong. A human-written application, by comparison, carries the irregularities of actual human drafting: varied syntax, imperfect keyword alignment, personality.

The result is a selection bias that no one designed and few employers have measured. Candidates who lean on generative tools may float to the top of the pile. Candidates who write their own materials — often the ones with deep, verifiable experience — may sink, not because they lack qualifications but because their prose does not resemble the machine's ideal.

Inc.'s framing is blunt: applicants need to know that the system rewards the system's own output.

Where does employer accountability sit?

The compliance and people-process implications stack up quickly:

• Adverse impact risk. If AI-favoring bias correlates with any protected characteristic — age, disability, access to AI tools — employers using these screeners inherit exposure under existing anti-discrimination frameworks.
• Validity of signals. Screening criteria that reward AI fluency rather than competence degrade the predictive value of the entire funnel.
• Audit obligations. Vendors rarely volunteer this weakness. HR teams should ask screening providers a direct question: how does your tool score AI-assisted versus human-written applications, and can you show the delta?

That last bullet is the practical takeaway. Vendor announcements celebrate accuracy and throughput; the question that matters here is whether the tool can even distinguish machine-written submissions, and whether it should.

Is the answer to fight AI with AI?

Some will read this and reach for detection software. That path has its own traps. AI-text detectors remain unreliable, and false positives would penalize candidates who happen to write in a polished, formula-driven style — a style the corporate world spent decades training them to adopt.

The more durable fix sits upstream: structured assessments, work samples and skills-based evaluation (hiring on demonstrated task performance rather than prose quality) shrink the surface area where AI polish can distort rankings. The less a hiring decision depends on how a document reads, the less it matters who — or what — wrote it.

Employers should also be honest with themselves about scale. As AI-assisted applications become the default rather than the exception, screening tools tuned to pre-2022 writing norms will drift further from reality. The quirk Inc. highlights today becomes the baseline condition of hiring tomorrow, and the employers who measure their own bias early will have the cleaner funnel.