AI Tools Are Breaking Corporate Hiring at Both Ends

AI Tools Are Breaking Corporate Hiring at Both Ends

Corporate hiring processes are facing a crisis of authenticity. Generative AI has made it trivially easy for candidates to produce polished résumés and perform well in structured interviews, regardless of their actual competence. The result is a systematic breakdown in how companies identify talent.

Research conducted by technologists Shraddha Sunil and Mudit Saraf, first reported in Harvard Business Review, reveals the scope of this problem. Over five months starting in July 2025, they interviewed 120 talent acquisition leaders across 87 companies and analyzed 6,380 recorded first-round screening sessions. Their findings show that early hiring filters are failing on both ends: résumés have lost signal value at the top of the funnel, while live video interviews are being compromised at the bottom.

Why it matters

When early screening mechanisms fail, companies don't just hire the wrong people—they systematically select for candidates who excel at gaming the hiring process rather than doing the job. This threatens talent density across organizations and can cost companies millions in bad hires, turnover, and lost productivity. For a company hiring 200 people annually, even small increases in screening errors create substantial direct costs before accounting for performance impacts.

The Résumé Problem

Candidates now generate tailored, keyword-optimized résumés in minutes using AI tools. But the issue runs deeper: a 2025 study by researchers at the University of Maryland, Ohio State, and the National University of Singapore found that AI screening tools show self-preference bias, favoring résumés that match their own output style. Candidates whose résumés resembled the evaluating model's style were 23% to 60% more likely to be shortlisted than equally qualified applicants who wrote their own materials.

One senior recruiter described the disconnect: "There is a growing gap between the candidate's written persona and their live presence. I'll see a cover letter that is poetic and a résumé that is flawlessly structured, but then the person on the video call struggles to explain their own bullet points."

Remote Interviews Under Pressure

Real-time AI assistance tools like Final Round AI and Parakeet now feed candidates scripted responses during video interviews without triggering traditional detection methods. The research found suspicious patterns varied dramatically by role type. For account executive positions, suspicion rates stayed below 10%. Mid-level technical roles reached nearly 40%. For new graduate software engineering positions, suspicious patterns appeared in almost 60% of sessions.

The researchers flagged sessions showing response-latency anomalies, sudden vocabulary shifts, and gaze patterns inconsistent with active recall. Sessions triggering multiple signals underwent manual review.

Gartner anticipates that by 2028, one in four candidate profiles will be partially or entirely fake. Companies like Google and McKinsey have responded by reintroducing in-person interviews for some candidates, a significant efficiency step backward that signals how seriously they view the threat.

Hidden Costs

Weak signals don't just allow wrong candidates through—they hide strong ones. The research showed that in adaptive, reasoning-based interviews, candidates with unremarkable résumés often outperformed credentialed peers. These candidates could explain tradeoffs clearly and reason through unfamiliar scenarios when conversations moved off-script.

As early filters weaken, recruiters increasingly retreat to known networks, abandoning open job postings. This quietly dismantles diversity by closing doors on non-traditional candidates who lack elite credentials but possess the right skills.

What Companies Must Do

Organizations need to separate authenticity verification from skills assessment. Early rounds should establish that candidates are real and capable of independent reasoning before deeper evaluation. First-round interviews must become genuine assessment layers using dynamic, adaptive questions that resist scripting.

Some companies are adapting. Meta recently began testing formats that allow candidates to use AI assistants during coding interviews, recognizing that the relevant question is whether candidates can use tools effectively and apply sound judgment, not whether they can work without them.

The details were first reported by Shraddha Sunil and Mudit Saraf in Harvard Business Review.