The Lazy Interview Is the Problem, Not the AI
Candidates using AI in interviews is not the crisis. The crisis is interviews so loose that an AI-assisted answer looks the same as a good one.
By August 2026, 38.5% of candidates in AI-powered interviews are being flagged for cheating behavior. Flag rates tripled in late 2025. In technical roles, it hits 48%. One-in-three hiring managers has interviewed someone who turned out to be a synthetic identity. Gartner predicts that by 2028, one-in-four candidate profiles globally will be fake.
Everyone is blaming AI.
They're wrong about what broke.
The Structure Was Already Missing
Here's what the data actually shows: 53% of employers lack standardized hiring practices. Nearly one-in-five interviewers receive zero training before sitting across from a candidate. Companies announced skills-based hiring, removed degree requirements from job postings, and called it transformation. The underlying evaluation loop — screening rubrics, structured interview question banks, consistent scoring criteria — stayed untouched.
So when AI showed up and made it trivially easy to feed perfect answers into a remote interview, it didn't break a functioning system. It exposed one that was already hollow.
The skills-based hiring movement is directionally correct. Skills predict job performance better than credential proxies. The research supports this clearly. But the movement's most prominent failure mode has also been documented clearly: removing degree requirements from a job posting is a single-day policy change. Rebuilding the rest of the hiring loop to actually evaluate skills is months of structural work that most companies never did.
They bought the headline and skipped the implementation.
What AI Actually Did
AI handed candidates an overlay tool. Real-time answer generation, fed directly from a screen the interviewer can't see. Predictable questions — the same behavioral prompts, the same technical scenarios companies have been running for years — get predictable AI answers.
One founder wrote about nearly hiring an AI candidate: two months, seven interviews, almost made the offer. The tells were subtle. Answers that were too frictionless. References that responded instantly, from thin email addresses with no LinkedIn history. Then, during a video call with a reference, she noticed the reference mirrored the candidate's exact speech patterns and mannerisms. When she pressed for verifiable HR contacts, he pushed to close the process faster.
Thirty minutes after she sent the rejection email, his LinkedIn profile disappeared. Every digital trace, gone simultaneously.
AI didn't create that vulnerability. The process did: remote interviews, unproctored assessments, references collected only from the candidate themselves, background checks scheduled at the end rather than the beginning.
The Actual Fix Isn't a Detector
The hiring teams beating AI fraud in 2026 share one thing: they stopped trying to catch cheaters and started designing interviews that make cheating useless.
The logic is clean. A fed answer is excellent. A fed answer under live follow-up falls apart.
"A genuine high performer can explain their thinking. A skillfisher can't." That's it. That's the whole strategy. Ask why. Change a constraint mid-problem. Ask for an edge case you invented thirty seconds ago. Push for the tradeoff they'd make if the requirements shifted. A model whispering answers into an overlay can produce output — it cannot produce the candidate's actual grasp of why the output is right.
This is what structured interviewing does when it's built properly: it forces reasoning into the open where it can't be faked. It also happens to be what skills-based hiring always required — not just removing the degree filter from the posting, but replacing it with something that actually measures the thing you care about.
The lazy version of structured interviewing is a checklist of behavioral questions read in the same order to every candidate. That's what AI can beat. The rigorous version adapts in real time, probes the edges, and scores reasoning rather than output quality. That's what AI can't beat.
Why Founders Get Exposed First
Startups move fast. Lean teams, compressed timelines, a bias toward trusting instinct. They're also least likely to have rebuilt hiring infrastructure after going remote, and most likely to skip steps under time pressure.
That combination — remote-first process, no structured rubric, reference checks at the close — is exactly the attack surface synthetic candidates exploit.
The fix isn't expensive. It's disciplined. Front-load identity verification. Run reference checks early, and independently — reach out directly to HR at the companies on the resume, not just the contacts the candidate provides. Use work samples with mandatory live walkthroughs. If you're fully remote, build in a small ritual: ask the candidate to hold three fingers in front of their face. Small friction, large signal.
The Candidate Who Vanished Wasn't the Problem
The story that should keep hiring teams up at night isn't the deepfake that got caught. It's the ones that didn't — the 60% of cheating attempts that go undetected, the new hires who arrived as someone their interview proved them to be, and proved nothing about who they actually are.
The answer isn't better detection software, though you should use it. The answer is an interview that doesn't need detecting — one where the real candidate is the only one who can pass it.
AI didn't break hiring. The lazy interview broke hiring. AI just made it impossible to ignore.
Jonathan Bouaziz is the founder of Umamy, an AI-powered hiring platform built to help founders and talent teams hire with structure, speed, and less bullshit.