The UV Light Test: How to Spot an AI-Polished CV Before It Costs You a Senior Hire
Written by: Matt Reaney | Reading time: 9 mins
If it feels right, it must be real. Right?
I gave my daughters a £10 note each recently instead of their usual pocket money app. Real money, in their hands. They loved it.
There was a story in the news that week about counterfeit notes circulating in the UK. And it occurred to me: how would they know? It looked right. Felt right. Had the right face on it.
That's the position most hiring managers are in right now with a senior technical CV.
How common is this?
Common enough that it's no longer an edge case.
Gartner predicts that by 2028, one in four candidate profiles globally will be fake. Their own survey of 3,000 job seekers found 6% openly admitted to interview fraud — posing as someone else, or having someone pose as them.
A Resume Genius survey found 17% of hiring managers had already encountered candidates using deepfake technology in interviews. Pindrop Security, a fraud detection company, posted a developer role as an experiment — 12.5% of applicants were using fake identities. A Blind survey from April 2025 found one in five US professionals had secretly used AI during a live interview.
Take whichever number you find most credible. The direction isn't in dispute.
This isn't opportunistic. It's an industry.
Counterfeiting isn't a UK problem — it's organised crime, operating in every major financial centre. New York, Singapore, Frankfurt, London. CV fraud in tech hiring now works the same way.
There are services for it. AI-generated CVs written to your exact job spec, for a fee. Interview answers scripted to your tech stack. Real-time overlay tools that feed responses to a candidate during a live coding round, invisible on the call. Candidates who've never touched a distributed system can describe one fluently.
Because they practised the words, not the work.
The difference between this and counterfeit currency is that your bank has a UV light. Most hiring processes don't.
Why your current filters no longer catch it
The CV is the first thing to go. Generative models write better CVs than most engineers do — perfect structure, quantified outcomes, keyword alignment. The document has stopped being evidence of anything except access to a chatbot.
Take-home tests are next. Outsourced or generated before submission, routinely. You're assessing someone's ability to commission work, not do it.
Standard technical interviews don't catch it either. A structured question set is a script, and scripts get prepped. The most uncomfortable finding in this space is that a large majority of candidates flagged for AI assistance still scored above the passing threshold. They would have been hired.
And the AI screening layer — the part people miss. Using AI to filter AI-written applications is symmetrical. Both sides converge on the same average, and the tool rewards the polished response over the real one. It doesn't detect the counterfeit. It grades the printing quality.
What actually catches it: pressure, from a peer
There is one thing an AI-polished candidate cannot survive, and it isn't a better test. It's an unscripted conversation with someone who has built the thing being discussed.
Not "rate yourself one to ten on Python." A real question, in real time, from a person who knows what the answer costs. Why that trade-off and not the other one? What broke? What did you try first that didn't work? Walk me through the decision you regret.
Prepped answers survive about ninety seconds of that. Real work survives indefinitely, because the person lived it.
This is not a recruiter skill. Pretending otherwise is how the industry got here. Recruiters failing in technical hiring is rarely about effort — it's about understanding. You cannot pressure-test a claim about inference latency if you've never had to own one.
The System Check
At reaney.ai we don't do the checking ourselves. Our Industry Advisory Panel does — active CTOs, CISOs and engineering leaders, currently building in the domain the role sits in. Every shortlisted candidate is put in a room with one of them, physical or virtual, and asked to defend the architecture of something they actually built.
That happens before a client sees a CV. Not after a first-stage interview. Before.
It's deliberately the opposite of automation. Automation converges to the average, and the average is exactly what a counterfeit is designed to look like. A peer under time pressure is the only instrument we've found that reliably separates the builder from the paper expert — and, just as importantly, recognises the outlier whose CV undersells them.
The cost of passing the note along
Volume recruiters aren't running this check. The economics forbid it — when you only earn on placement, you can't afford to hold anything up to the light. So the note gets passed along and hoped about.
You're the one left holding it. A failed probation, three months of your engineering team's time, a second search starting from zero.
My daughters knew that tenner was real the moment they held it.
Your next senior hire should feel exactly the same. Real is rare now — and rare is worth paying for upfront.
FAQ
How long should it take to hire a senior engineering lead?
With a properly scoped brief and peer-led vetting, a verified shortlist inside 30 days is achievable. Searches that run longer usually mean the brief wasn't right to begin with — not that the talent isn't there.
Is retained search worth it for one hire? Compare it to the downside rather than to a contingent fee. Standard estimates put the cost of a failed senior hire at 50–200% of salary once lost productivity and a second search are included.
What is the 30-Day Technical Shortlist?
A defined shortlist of candidates verified by our Industry Advisory Panel of active CTOs and engineering leaders, delivered within thirty days of briefing. Not volume. Verified.
Matt Reaney has been building AI and machine learning teams since 2013. Book a search briefing today.