A few months ago, I went through an interview process that ended with some feedback I honestly didn't expect. The recruiter told me that the interviewers thought I might have been using AI to answer some of the questions.
My first reaction was basically:
"Wait... what?"
Not because I was offended, but because I found it genuinely funny.
The process had started with a take-home assignment. I built a small SpaceX-themed website that consumed an API and put together what I thought was a reasonably clean React application.
Nothing particularly revolutionary. Just the kind of thing I've been doing for years.
Then came the technical interview.
They asked me about React, including things like the Virtual DOM, React Query, caching, and a bunch of other topics that are probably very useful to know if you're building a React framework from scratch.
They also asked about Backend-for-Frontend architectures and NestJS.
I didn't implement a BFF in the take-home assignment. They were simply questions about how I would approach it, why I would use one, and the trade-offs involved.
I answered based on how I've actually worked with these technologies.
Apparently, some of those answers sounded too good.
So good, in fact, that they wondered whether AI was helping me.
And that's where things get interesting.
What are we actually trying to measure?
I completely understand why interviewers are cautious about AI.
Take-home assignments are particularly difficult now. If you ask someone to build a small application over a weekend, there is a very good chance that AI will be involved somewhere in the process.
And honestly, I don't think that's necessarily a bad thing.
Developers have always used tools.
We use IDEs, linters, documentation, Stack Overflow, GitHub, autocomplete, debuggers, profiling tools, and increasingly, AI assistants.
The interesting question isn't really:
"Did you use AI?"
It's:
"Do you understand what you built?"
If I use an AI assistant to help me remember the exact syntax of something, that's very different from having AI build an entire system that I cannot explain.
The same applies during an interview.
If someone asks me how React Query handles caching, I don't need AI to tell me the answer. I've used the library. I understand why caching exists, what problems it solves, and what trade-offs come with it.
Likewise, if someone asks me about a BFF architecture, I don't need to memorize a textbook definition. I've spent enough time building distributed systems to reason about where a BFF makes sense, where it doesn't, and what problems it introduces.
That's the part of engineering experience that is difficult to capture with trivia questions.
AI is changing the interview too
There is an interesting paradox happening right now.
AI is changing how software engineers work, but many technical interviews are still designed around evaluating whether someone can produce answers without assistance.
At the same time, interviewers now have to consider whether an answer that sounds unusually polished might have been generated by AI.
So we end up in a strange situation where:
If you answer poorly, you might not get the job.
If you answer really well, they might wonder whether AI helped you.
Maybe that's a sign that the way we evaluate engineers needs to evolve.
I'd much rather be asked to reason through a real problem, explain my decisions, challenge an assumption, debug something, or discuss the trade-offs of a design than recite implementation details from memory.
Because that's much closer to what engineering actually looks like.
And yes, AI is going to be part of that engineering process.
The ability to use it effectively while still understanding, reviewing, and owning the result is becoming a skill in itself.
And then, a plot twist that actually fit the story
After all those interviews, take-home assignments, and the unexpected AI accusation, I ended up back where I started: at bol.com, this time as a Senior Full Stack Engineer on the Ad Advice team.
But here's the part that actually connects to everything above.
What made me want to come back wasn't nostalgia. It was realizing, through that whole process, what I actually value in a place I work: an environment where the conversation is "walk me through how you'd solve this" instead of "did a machine help you sound smart." Three years there, a stint on distributed systems at Tesla, and a handful of interviews later, that's the kind of engineering culture I was looking for again.
So maybe the real plot twist isn't that they thought I was using AI.
It's that the accusation ended up teaching me exactly what I should be optimizing for when I pick where to work next.