There are two ways to get interview integrity wrong. The first is to do nothing and let AI tools quietly answer for the candidate. The second is to over-correct — lock everything down, treat every applicant as a suspect, and damage the experience for the honest majority. The goal of a good process is to avoid both.

The pressure is real. Gartner found that 39% of candidates used AI somewhere in the application process and 6% admitted to interview fraud, while a Blind survey reported one in five professionals secretly using AI during interviews. Cheating is common enough to justify doing something about it.

Why the interrogation approach backfires

Candidates notice how they are treated, and it follows them. A long-running meta-analysis of nearly 49,000 applicants found that positive perceptions of the selection process strongly raise the intent to accept an offer and recommend the employer — and negative perceptions cut both. Heavy surveillance erodes that trust: only 26% of candidates trust AI to evaluate them fairly, and 62% are more likely to apply when a process includes real human interaction.

Why detectors alone are dangerous

The instinct to bolt on an "AI detector" runs into hard math. When real cheating is relatively rare, false positives outnumber true positives even with a very accurate test. The tools themselves are unreliable: OpenAI discontinued its own AI-text classifier for low accuracy (it caught 26% of AI text and false-flagged 9% of human text), and a Stanford study found GPT detectors wrongly flagged 61% of essays written by non-native English speakers as AI. A single flag is never proof — and treating it as one quietly builds bias into your funnel.

A process that is fair and still catches cheats

The fix is not more suspicion. It is structure.

The balance

Cheating is real, so detection earns its place — a bad hire still costs employers about $14,900 on average. But detectors fail the base-rate test and discriminate, so the only defensible design is a structured, transparent, consent-first process that weighs several signals and lets a human make the call. That protects the honest candidate and still catches the one reading from a hidden screen.

That is the read Trueyy is built to give you: the small tells, tied to the moment they happen, so your team acts on evidence instead of suspicion.

Sources