How to Answer "Do You Use AI at Work?" in an Interview, and How to Practise Spotting AI Mistakes

By Pooja Goenka · Published 2026-10-14

If an interviewer asks whether you use ChatGPT or other AI tools, say yes if you do, name the tool and the task, and spend most of the answer on how you checked what it gave you. Interviewers are listening for judgement, so one specific story about catching an AI mistake beats a list of tools.

In short:

  • Be honest. A "no" that is not true is easy to expose with one follow-up question.
  • Use three parts: the tool, the task, and the check.
  • The check is the part that earns marks, so give it the most time.
  • Practise finding mistakes in AI answers before the interview, because some interviews now test it directly.

What the interviewer is really asking

The question sounds like a yes-or-no about tools, though what the interviewer usually wants to know is whether you can be trusted with them.

Most developers already use these tools. In the 2025 Stack Overflow Developer Survey, 84% of respondents said they use or plan to use AI tools in their development process. The same survey found more developers distrust the accuracy of AI output (46%) than trust it (33%), and the most common frustration, cited by 66%, was "AI solutions that are almost right, but not quite."

That last number explains the question. An interviewer knows you have probably used ChatGPT. What they cannot tell from your résumé is whether you notice when it is almost right. A candidate who pastes AI output into a project without reading it creates work for the team. A candidate who uses it and checks it saves the team time.

So the question is really three questions. Do you use these tools? What for? And do you catch them when they are wrong?

A three-part answer: tool, task, check

Structure the answer like this:

  1. Tool. Name it plainly: ChatGPT, GitHub Copilot, Gemini, Claude, whatever you actually use. One or two is enough.
  2. Task. Say what you used it for, specifically. "Writing the first draft of SQL queries for my mini project" is good. "Coding" is too vague to mean anything.
  3. Check. Say how you verified the output, and if you can, describe one time it was wrong and how you found out.

Give the tool and task a sentence each and the check most of your time. If your answer runs for a minute, at least half of it should be the check.

Three sample answers for freshers

These are templates. Swap in your own tools, projects and mistakes. Do not make up an incident; a follow-up question will expose it within a minute.

When you used it for code

"Yes. I used GitHub Copilot in my final-year project, mostly for boilerplate like Flask routes and form validation. I read every suggestion before accepting it, and I wrote tests for the validation logic myself. One suggestion used a date comparison that treated the end date as exclusive, so bookings on the last day failed. My test caught it, and since then I always test edge values when Copilot writes a comparison."

When you used it for learning

"Yes, mainly ChatGPT, to explain concepts while I was learning SQL. I learned the hard way to check its claims. It once told me COUNT(column) and COUNT(*) give the same result, and when I ran both on my own table with NULLs in it, they did not. Now I treat its explanations as a starting point and test anything I plan to rely on."

If you want more examples of SQL answers that sound right and are not, I collected eight common ones, each run in a real database.

When your college restricted it

"For coursework, our college did not allow AI tools in submissions, so I did not use them there. For my own side project, I used ChatGPT to help debug error messages. I always reproduced the fix myself and checked the official docs before keeping it, because it sometimes suggested library functions that did not exist in the version I was using."

This one shows something interviewers also want to see: that you follow the rules of the place you are in.

Follow-up questions to expect

Prepare for these, because a good interviewer will go one level deeper:

  • "How do you decide when not to use AI?" A reasonable answer mentions sensitive data, code you do not understand yet, and situations where checking the answer would take longer than writing it yourself.
  • "Tell me about a time it was wrong." Have one real example ready, with what the mistake was and how you found it.
  • "How would you check this answer?" Some interviewers will show you an AI-written answer on screen and ask you to review it, and that round is worth practising on its own.
  • "Would you paste company code into a public chatbot?" The safe answer is that you would follow the company's policy, and that you would not paste confidential code or data into any tool the company has not approved.

How to practise catching AI mistakes

Reviewing an AI answer is a skill you can drill. The method I recommend is to read the answer line by line, name the rule each line depends on, and ask what input would make that line false. I worked through it with two examples in how to practise spotting an AI's wrong answer.

A few habits help:

  • Be suspicious of lines that sound most confident, especially ones with "always", "never" or a specific number.
  • Run the code when you can. A thirty-second test settles most arguments.
  • Look for the claim that is almost right: the right function with the wrong default, or the right concept applied to the wrong case.
  • Do not flag lines you are unsure about just to be safe. Marking a correct line as wrong is also a mistake, and good interviewers count it as one.

For GenAI roles, the same skill applies to concepts. I collected ten GenAI interview questions where the confident answer is wrong, and each one is a good drill.

Free practice on LogicWiz

We built two free ways to practise this.

LogicWiz Interviews has AI mock interviews with 100 curated questions across 16 tracks, including backend, SQL, machine learning and behavioural. Its spot-the-mistake mode has the AI write a confident, well-structured answer with one planted technical error, and you have to find it. Confidently flagging something that was correct counts as a miss, the same way it would with a careful interviewer. Every session ends with a scored report, and your practice is private and never shared with an employer. You get 3 AI practice sessions a day with a free account.

LogicWiz Arena has a game called Spot Aria's Mistake. Aria answers a technical question in five lines and exactly one line is wrong. You tap the wrong line, then pick why it is wrong. A game is five rounds and takes about five minutes, across eight tracks from AI and Python to SQL, cloud and DSA. A practice round needs no account, and every player gets a certificate. There is more on how it works in our post about the game.

Common questions

Is it bad to admit using ChatGPT in an interview?

No. Most developers use AI tools, and interviewers know it. What hurts you is using it without checking, so lead with how you verify its output.

What if my college bans AI tools?

Say so, and say you followed the rule for coursework. If you used AI on personal projects, describe that use and how you checked it. Following the rules where you are is a point in your favour.

How do I practise this for free?

Review AI answers line by line and test the claims yourself. LogicWiz Interviews has a free spot-the-mistake mode (3 sessions a day with a free account), and Spot Aria's Mistake on LogicWiz Arena is a free five-round game.

Will the interviewer test me on an AI answer?

Some will. Some technical interviews now include a round where you review an AI-written answer and find what is wrong, so practise that as well as the verbal question.

Should I list AI tools on my résumé?

Only the ones you have used on real work, and preferably tied to a project. "Used GitHub Copilot in a Flask project; wrote tests for generated code" says more than a skills line that just reads "ChatGPT".

Back to the blog