The AI Penalty Is Real: Coding Interviews Just Got Weirder
Coding interviews are about to get weirder.
I’ve done no less than 10 interviews in the last 12 months. TBH, they’re mostly the same.
- Live code something with React.
- Maybe a LeetCode medium.
- Awkwardly turn off Co-Pilot or Cursor so you don’t complete their challenge in seconds.
But recently, something changed.
The unexpected AI penalty.
Many companies explicitly ban you from using AI tools during the interview. If you’re caught, it could ruin your reputation and get you on their do-not-hire list for life.
Some companies are opting for in-person interviews to deter cheating.
Some companies are going in the exact opposite direction.
Check out this recent job posting for a start-up in LA:
If you’re a developer who is avoiding AI tools in order to “hang on to your coding skills” — this may actually hurt you in your next interview.
My last interview reflected this new reality.
“It’s fine. Keep Cursor on.”
My most recent interview involved a coding challenge to create a full stack NextJS app with an API route that did some string manipulation.
I shared my screen and opened up my preferred editor, Cursor.
Oops.
“Oh sorry, I can switch to VS Code and turn off AI tools for this exercise” I said.
The CTO paused briefly.
“No — it’s fine, keep it on. We want to see how you use AI to approach the challenge. We want you to tackle this the same way you would as if you were working.”
To say I was shocked is an understatement.
It makes sense though.
AI is an integral part of what we do as coders in 2025. I used to be asked how I found answers to problems using Google. Now I was going to expose how I use Cursor to my interviewers.
This did not make my interview any easier. It may have been more difficult actually.
Why didn’t you use the approach the AI gave you?
Blindly copying code from an AI tool is as bad as blindly copying from Stack Overflow.
Actually, it’s worse.
AI can generate slop at a speed we can’t really fathom. If you honestly believe the code generated by AI is THE best approach — you might actually want to go old school until you can use your AI tool the way it works best: as your assistant, not your boss.
During the fairly simple challenge, I used AI to scaffold the front end in a single prompt.
The backend code it generated however, was un-necessarily complex and bloated.
I rejected the response it gave me and opted to code some of it on my own.
The CTO asked why.
“Any reason you rejected the code it gave you? It looked decent.”
It would work for sure, I explained.
However — did we really need to use regex? Would a map work better than an object? Can we decompose this massive function into a few smaller functions for re-use?
Overall he agreed with my approach and it led to good discussion.
That’s the point.
He’s not hiring an AI agent, he’s hiring a human.
For now.
If I had copied and pasted the output from Cursor, I’m sure it would work with enough prompting but if you’re not considering maintainability, simplicity and test-ability, then you’re a slave to your tool instead of the other way around.
Defending my decisions to use or not use AI-generated code during the interview was much more difficult than I anticipated.
Knowing the fundamentals of TypeScript, Javascript and some basic coding principles like the rule of Single Responsibility or Open-Closed principle were useful to explain my thought process and create trust with the interviewers.
Know your tools, don’t forget your fundamentals.
In 2025, knowing how to use AI tools like Cursor is part of the job. But using them blindly is a fast track to getting exposed in an interview.
I should also mention that out of ~30 interviews this team conducted, I got the offer.
I highly doubt I was the best at using Cursor. This low pass rate may say more about the quality of senior engineers out there than anything.
Here’s what I’d recommend if you want to stand out in this new era of interviewing:
- Use your AI tools and be intentional. Don’t just prompt. Edit. Refactor. Reject suggestions that are bloated or don’t align with best practices.
- Brush up on your fundamentals because knowing why you’re rejecting AI output matters more than just spotting that it “feels wrong.”
- Practice defending your code — explain your choices. Interviewers are listening for reasoning, not just results.
Your next interview might come down to you vs. someone who also uses AI and knows when to say no to it.
Change is exciting.
Change is scary.
We fail, we learn and we adapt.
As always I hope you found this useful and good luck if you’re preparing for interviews.
At Parsity we work with a handful of career changers who want to break into tech with 1 on 1 mentorship from senior+ engineers. Apply here.
