Banning AI from Tech Interviews Is the Wrong Move

Bybrjorgedavid

There’s a trend happening in hiring right now that deserves pushback.

Companies are banning AI tools from technical interviews. No Copilot. No ChatGPT. No Cursor. Candidate opens a blank editor and writes code from memory, under time pressure, in front of a stranger on a Zoom call.

The stated reason is fairness and signal quality. The actual effect is a hiring process that evaluates a skill set that no longer reflects how software gets built.

That’s a problem worth talking about.

The Stat That Makes This Argument Short

According to JetBrains’ January 2026 developer survey, 90% of developers regularly use at least one AI tool at work for coding and development tasks.

Not occasionally. Not as an experiment. Regularly. As part of how they do their job.

That number has only grown since January. We’re writing this in May.

If AI-assisted development is the standard — and at 90% adoption, it clearly is — then an interview process that strips away AI isn’t testing how candidates work. It’s testing how they work under artificial constraints that don’t exist in the role they’re being hired for.

That’s not signal. That’s noise.

The LeetCode Problem Predates AI

To be fair, the technical interview was already broken before AI made it obvious.

LeetCode launched in 2015 and quickly became the default benchmark for “objectively” assessing engineering ability. The logic was appealing: standardized problems, measurable outcomes, no bias. Clean signal.

The reality was different. What LeetCode actually measures is whether a candidate has spent weeks grinding algorithm problems on LeetCode. That’s it.

I’ve spoken with engineers at every level who describe the same experience: prepare intensively for weeks, learn the patterns, pass the interview, get the job — and then never use any of it again. The work was completely disconnected from the preparation.

The interview was testing practice, not capability.

And now, in 2026, with AI able to solve any LeetCode-style problem in seconds, the signal is even weaker. You’re not testing engineering skill. You’re testing whether someone memorized an approach to a problem class that AI renders trivial.

What a Technical Interview Should Actually Measure

Here’s the framing that cuts through the debate:

Modern engineering is: good judgment + AI to move faster. Judgment is what you’re hiring for.

An engineer’s value in 2026 isn’t their ability to implement a binary search from memory. It’s their ability to look at a system and ask the right questions. To understand trade-offs. To know when a simple solution beats an elegant one. To catch what the AI missed. To push back when the requirements are wrong.

That’s not something you can LeetCode your way to. And it’s exactly what a well-designed interview can reveal — if the interview is designed to look for it.

The question shouldn’t be “can this candidate write a working solution without tools?” It should be: “does this candidate demonstrate good judgment when using the tools they’ll use every day?”

Those are very different assessments. Only one of them tells you something useful about the hire.

The Inconsistency That Should Bother Engineering Leaders

Here’s the part that’s hardest to defend: many of the companies banning AI from interviews are simultaneously pushing for more AI adoption internally. They have AI usage goals. They’re paying for GitHub Copilot seats. They’re measuring AI-assisted productivity gains.

And then they’re hiring engineers by evaluating whether they can work without AI.

The candidate who passes your no-AI interview is being asked to demonstrate something you don’t actually want them to do on the job. That’s an alignment problem. It’s testing for a skill you’re actively trying to move away from.

The interview process should reflect the work. If the work involves AI — and at 90% adoption, it does — the interview should too.

What Should Change

The goal of a technical interview hasn’t changed: assess whether this person can do the job well. What’s changed is what the job looks like.

A 2026 engineering interview that’s worth something should evaluate:

System thinking over syntax recall. Can the candidate reason about trade-offs? Do they ask clarifying questions before jumping to implementation? Do they consider failure modes?

Code ownership with AI assistance. Give them a realistic task, let them use AI, and watch how they engage with the output. Do they read it carefully? Do they question it? Do they catch the bug the model introduced? That last part, catching what AI misses, is one of the most valuable skills in modern engineering.

Debugging and reasoning. Take an existing piece of code with a problem. Can the candidate understand it, reason about why it’s failing, and fix it — with or without AI? This is much closer to real work than writing a solution from scratch.

Communication and judgment under real constraints. Can they explain their decisions? Can they articulate why they chose one approach over another? Can they push back on a requirement that doesn’t make sense?

None of these require removing AI. All of them reveal whether the candidate is actually good at engineering.

The Direction This Is Heading

Here’s a prediction that feels safe to make: by the end of 2026, the companies still running no-AI LeetCode interviews will be at a disadvantage in hiring. Not because candidates will refuse to participate — some will, but that’s not the main issue.

The issue is that they’ll be optimizing for the wrong candidate. The engineer who’s great at grinding algorithm problems under no-tool constraints and the engineer who’s great at building real systems with modern tools are not the same person. And increasingly, the second type is the one worth hiring.

The interview process is also a signal to candidates about how a company thinks about engineering. A company that bans AI from interviews is, whether it intends to or not, signaling that it doesn’t fully understand how software gets built right now. That signal reaches the candidates you most want to hire.

The Bottom Line

Removing AI from technical interviews isn’t rigorous. It’s nostalgic.

The craft of engineering has always involved tools — compilers, IDEs, Stack Overflow, documentation, frameworks built by other people. Nobody argues that using an IDE is cheating. Nobody makes candidates write code in a terminal with no syntax highlighting to test whether they “really” know the language.

AI is the next tool in that lineage. The engineers who use it well are the ones building the best software right now. Hiring processes that ignore this reality aren’t protecting signal quality. They’re testing for a past that no longer exists.

The interview should reflect the job. In 2026, the job involves AI.

References

Jorge David has been in technology since 2004, working across IT infrastructure and software development with Java, Kotlin, Spring Boot, and Kafka. Dev AI Tools covers honest perspectives on AI and the engineers who use it. 

Similar Posts

Leave a Reply

Your email address will not be published. Required fields are marked *