We've started seeing CVs that were obviously written by ChatGPT. Here's how we screen for what's actually behind them.

The first one we caught was for a financial manager role. The CV used the phrase "leveraging cross-functional synergies" three times in different sections, and the cover letter opened with "In today's rapidly evolving business landscape." Both phrases scored zero on the question they were supposed to answer.

This is going to be common in South African recruitment over the next two years. ChatGPT (and Claude, and Gemini) makes it trivial to write a CV that hits every keyword in the job description, sounds confident, and tells you almost nothing about whether the candidate can actually do the work. The keyword-matching tools that most ATS systems rely on were already gameable. AI makes them gameable at scale.

Here's what we're doing on our end.

We ignore CVs as the primary signal. We never did love them — they're a self-edited marketing document at the best of times. They're worse now. What we look for is what shows up on the phone screen. The phone screen is where you find out whether the person can describe their own work in their own words. AI helps you write a CV. It doesn't help you describe what you actually did in a project last week.

We ask for examples that aren't in the CV. "Tell me about a time you had to fire someone and what you said in the first sixty seconds of the conversation." "Walk me through how you structured the spreadsheet you used to track that project." "What did your manager push back on when you proposed that change?" These are questions a generated CV can't answer because the CV didn't know the answers. If the candidate hesitates and then constructs an answer that sounds suspiciously like the CV's framing, that's information.

And we've stopped treating CV polish as evidence of competence. A beautifully formatted CV with strong action verbs used to mean someone who'd been trained to write strong CVs, which usually correlated with someone who'd had professional jobs. Now it means someone who has thirty seconds and a free AI account. The correlation broke. We've stopped using it.

A small admission. We use AI ourselves. I use it to help draft job descriptions when a client gives me a brief in three sentences. Monique uses it to summarise reference call notes and to draft candidate-experience emails. It's a useful tool. The problem isn't the tool. The problem is when the tool is the only thing between the candidate and the hiring decision.

If you're a hiring manager wondering whether AI-written CVs are going to fool you, here's the honest answer. Yes, they probably will, on the page. They won't fool you in a structured forty-minute conversation with the candidate. Which is what you were going to do anyway.

— Kern