How to Interview People When AI is in the Room (A Survival Guide)
TLDR: Interviewed someone using AI in real-time. Here’s how to spot it (eye movements, timing patterns, essay-structured answers) and tactics that actually work (funnel questions, rapid-fire, asking for opinions).
I recently interviewed someone who was 100% using an LLM in real-time.
I asked about their tech stack. Got back something about modern software development principles and comprehensive suites of technologies.
Dude. Just say React.
But the real kicker? When I asked “are you using AI right now?” they used AI to answer that too.
I’m not even mad. I’m impressed by the commitment.
This is only going to get more common. So instead of being mad about it, let’s figure out how to deal with it.
How to Tell When Someone’s Using AI
The eye test
Biggest tell: Real people’s eyes move around when thinking – up, to the side, away from camera. AI people? Stare directly at their screen. No movement. Just locked in for 15 seconds. Then perfect answer.
The timing pattern
Any question gets the same delay. Simple question? 15 seconds. Complex question? 15 seconds. Like they’re buffering.
Real people answer “What’s your notice period?” instantly. But take time on “Tell me about your biggest failure.” AI people treat everything the same.
The reliance problem
Here’s what I’ve noticed: Once they start, they can’t stop.
It’s like they got into the habit of reading off the page and now can’t look away. You’re asking technical questions, they’re in full robot mode – staring at screen, 15-second pauses, essay answers.
Then you switch to “tell me about a time you disagreed with your team” and… same robot. They’re stuck. Can’t change horses mid-race.
The tool becomes a crutch for EVERYTHING, including questions where it makes them sound worse. I’ve watched people route “how was your weekend?” through an LLM. That’s not optimization, that’s dependency.
And when they do use it for everything? Here’s what it sounds like:
They write instead of talk
Real human: “Yeah the API was broken, took forever to debug”
AI human: “Leveraging cross-functional synergies to drive compelling value propositions across the technology stack”
And it’s always structured like an essay – “That’s a great question,” three supporting points, “In summary…” Nobody talks like that. That’s how we write college papers at 2am.
They can’t handle simple questions
Ask “Did you like it?” and get back a paragraph about organizational culture and professional development.
Just say yes or no!
When you call them out, even that’s scripted
Ask “Are you using AI right now?” and a real person gets flustered or defensive or laughs it off.
AI person: Something about appreciating directness and leveraging various technologies (seriously, that was the answer).
Interview Tactics That Actually Work
So you’ve spotted the tells. Now what? Here are tactics that actually work to screen these people out:
The funnel approach – start big, get tiny
Start broad, then narrow down fast:
- Tell me about your last project
- What tech did you use?
- Why that one?
- Did you like it?
- What annoyed you about it?
- React or Vue?
- Why?
Each question gets simpler. Real people adjust their answers accordingly.
Example: Question: “Did you like it?” Real person: “Yeah, mostly. The state management got messy though.” AI person: “This experience provided valuable insights into organizational culture and professional development opportunities within a dynamic technical environment…”
Rapid fire questions
“Quick, first thing that comes to mind – best bug you ever fixed?”
Real person: “Oh man, okay so—” laughs “—there was this one time…” and they light up
AI person: 15 second pause “In my experience, effective debugging requires a systematic approach with appropriate consideration of…”
Come on.
Ask for opinions, not facts
Don’t ask what they know. Ask what they think.
“What’s overrated in your field right now?” “What’s your hot take on [technology]?” “What hill will you die on professionally?”
Opinions are messy and personal. Can’t polish those without losing what makes them real.
Ask meta questions
“What question are you hoping I don’t ask?” “What should I have asked that I didn’t?” “If you were me, what would you want to know?”
These need actual presence and self-awareness. Can’t really script them.
Real Talk: Using AI to Prep is Smart
Look, I’m not anti-AI. Using it to prepare is actually smart. Let’s be clear about the line:
Before the interview? Hell yeah:
- Research the company
- Practice answers and get feedback
- Review rusty concepts
- Prep good questions
During the interview? Nah:
- Real-time copy-paste answers
- Needing AI to have basic conversation
- Corporate speak that sounds like “pursuant to established protocols”
The test: If the job requires talking to people in real-time and you need AI for that in the interview… you can’t actually do the job.
What’s Coming Next
This is going to get harder. AI will get better at sounding human. Deepfaked video. Cloned voices. We’re heading toward a point where we might not be able to trust we’re talking to a real person at all.
So what’s the move? More in-person finals, pair programming, and asking questions AI can’t prep for. Long-term, we need to shift from testing “what do you know” to “how do you think” – because AI already knows everything. The interview needs to be about real-time problem solving, actually communicating with humans, and whether they’d be good to work with. That stuff’s harder to fake. For now anyway.
The Bottom Line
The person I interviewed? Solid resume. Real experience. (In hindsight, probably AI-written too – every keyword from the job description was in there.)
But they bombed because they let AI do the talking. Every answer sounded like HR wrote it.
The irony is perfect – they used AI to seem more competent and ended up seeming less human.
And honestly? That’s the real test now. Not “can you answer the question” but “can you sound like an actual person while doing it?”
Game’s changed. Time to get creative.
Written with an LLM and opinions are my own 😉
