Righty ho. This week's topic came from something I read that made me sit up properly. Not in a doom-and-gloomy way. More in a "well, that changes a few things" way. We're going to talk about what your candidates are doing before they walk into your interviews.
Your candidates are coached by AI....Cool!
70% of job seekers now use AI to research companies, draft cover letters, and prepare interview talking points. That's not a prediction. That's the current number.
But what caught my attention wasn't people using ChatGPT to tidy up their CV. That's been happening for years. What caught me was how far ahead of us some candidates have gone.
Noam Segal, the Community Research Lead at Lenny Rachitsky's newsletter (Lenny's Newsletter, 1M+ subscribers, one of the biggest product and tech newsletters going), interviewed over 30 tech professionals about how they use AI in the interview process. The findings were published in February 2026 and they're worth your time.
Here's what the best candidates are now doing.
They're building entire prep systems.
The candidates Noam spoke to aren't asking ChatGPT "how do I answer a competency question." They're building structured AI workflows that cover the full interview lifecycle:
Feeding their actual interview transcripts (recorded via Granola, Zoom, or Google Meet) into Claude and getting line-by-line feedback on what they said versus what they should have said.
Pasting a job description alongside their CV and having AI surface the exact gaps a hiring manager would flag, before the review happens.
Simulating full mock interviews in advance, testing which career stories land, and refining them before the real thing.
Building "story banks" of 8-15 indexed career stories, tagged by skill and impact, with rehearsed opening lines and rapid retrieval drills.
One candidate who hadn't interviewed in eight years landed a senior role at Anthropic with two weeks of prep. He credited his AI workflows as the primary reason.
Noam took everything he learned from these 30+ conversations and built a free, open-source AI interview coach using Claude Code. It scores candidate answers on five dimensions (substance, structure, relevance, credibility, differentiation), runs company-specific prep with predicted questions, does full mock interviews, and even coaches salary negotiation with exact scripts. It's on GitHub for anyone to use.
So what does this mean for recruiters?
Firstly I think this is a great development but heres three practical things.
- Your questions need to go deeper.
If candidates can predict your interview questions using AI (and they can), your questions are too generic. "Tell me about a time you dealt with a difficult stakeholder" is now a rehearsed answer with a polished STAR structure before the candidate walks in.
Go deeper. Ask follow-ups that go past the rehearsed layer. "What specifically went wrong?" "What would you do differently now?" "What did you learn about yourself?" That's where real experience shows up, because AI can prepare the headline answer but it can't fabricate the nuance underneath.
- A polished answer is no longer a reliable signal of depth.
This is the calibration challenge. A candidate with two weeks of AI-assisted prep can now present as smoothly as someone with months of genuine preparation. That's good news for talent quality overall, but it means interviewers need to probe for specifics rather than being impressed by structure.
The data backs this up. 77% of hiring managers say they can now detect AI-assisted candidates, up from 53% two years ago. But detection isn't the point. The point is that preparation quality has gone up across the board, and your interview process needs to assess capability, not just presentation.
- Build your own feedback loop.
Candidates are now analysing their own interview performance with AI. They record, transcribe, and review. They know which answers were weak. They iterate. By their third interview, they're significantly sharper.
Meanwhile, most recruiting teams have no structured feedback loop on their own interview quality. We don't record our debriefs. We don't track which questions surface the best signal. We don't know which interviewers are calibrated and which ones aren't.
If candidates are using AI to get better at interviewing, we should be using AI to get better at assessing. Feed your interview notes into Claude. Ask it to find patterns. Which questions consistently surface strong signal? Which ones are a waste of time? Which interviewers flag the same concerns every time regardless of who they're speaking to?
The feedback loop has flipped. Candidates have one now. Most recruiters still don't.
So....
This isn't about candidates cheating. Using AI to prepare is resourceful, not dishonest. BUT..is your interview process designed for a world where prep doesn't indicate quality?
HUGE Credit where it's due
This issue was heavily inspired by research published by Noam Segal, Community Research Lead at Lenny Rachitsky's Newsletter. Noam interviewed 30+ tech professionals, synthesised their AI interview techniques, and built a free Claude Code-based interview coach available at: https://github.com/noamseg/interview-coach-skill
The original article was published on 24 February 2026: https://www.lennysnewsletter.com/p/how-to-use-ai-in-your-next-job-interview
Additional data sourced from: https://resumegenius.com/blog/job-hunting/hiring-insights-report and https://www.demandsage.com/ai-recruitment-statistics/ and https://www.socialtalent.com/blog/interviewing/ai-candidates-are-already-here-your-interviewers-arent-ready
If you want to go further down this rabbit hole, check out the Prompt Mates community a Slack group run by Jason Miller and Joe Atkinson where recruiters are building AI tools, sharing workflows, and generally being the sort of people who make this industry better.
Well... that's issue sixty eight! If you have any questions about it, or any feedback on this issue of The Data Driven Recruiter, grab me on LinkedIn for a chat.
I'll see you next week!