The AI-assisted coding interview: what changes when AI is allowed
For years the rule in coding interviews was simple: no AI. A growing number of companies now run rounds where the AI is allowed, or even required. The skill being tested changes with it. These rounds are not about whether you can write a binary search from memory. They test whether you can get a correct, maintainable result out of an AI assistant under time pressure, and whether you notice when it is wrong.
Who is already doing it
- Canva was the most explicit. In June 2025 its engineering blog announced "Yes, you can use AI in our interviews. In fact, we insist." Backend, machine learning and frontend candidates are expected to use tools like Copilot, Cursor or Claude. The new AI-assisted coding round replaced the old computer science fundamentals interview, where code had to be written without AI. Canva gave its reason: almost half of its frontend and backend engineers use an AI coding assistant every day.
- Meta has been building an AI-enabled coding interview. In 2025, 404 Media reported on an internal post recruiting employees as mock candidates for it, and Meta confirmed the effort to Business Insider. The format was described as experimental, so treat it as a pilot rather than a company-wide rule.
- CodeSignal sells the format to employers as AI-Assisted Coding Assessments and Interviews. Candidates work with a built-in assistant called Cosmo, in either a full co-pilot mode or a lighter guided mode. The employer receives a full transcript of the conversation with the AI and a replay of the session.
The policy is set by each company, not by the industry. Amazon, for example, was reported to tell recruiters to disqualify candidates caught using AI in an interview. The same platform can run an AI-allowed round for one employer and a strict no-AI round for the next.
What these rounds actually grade
Canva lists what its interviewers look for, and it lines up with what the other formats measure:
- Breaking down an ambiguous problem before writing any prompt. Asking clarifying questions counts in your favor.
- Knowing when to use the AI and when not to. Handing over every step is a negative signal, not a positive one.
- Reviewing AI output critically. Finding the bug in generated code and fixing it is often the point of the round.
- Bringing the result to production quality: edge cases, naming, tests, structure.
Your prompts are part of your answer. When the employer gets a transcript, a prompt like "solve this" reads very differently from one that states the constraints and asks for a specific piece.
How the problems change
If the AI is allowed, a LeetCode puzzle stops being a useful test, because the assistant solves it in one shot. So the problems get bigger and vaguer. Canva says its tasks "can't be solved with a single prompt". Its example is not Conway's Game of Life but "build a control system for managing aircraft takeoffs and landings at a busy airport". Expect a task with several parts, unclear requirements and room for design decisions.
How to prepare
- Read the invitation carefully. It tells you which assistant is allowed. Some rounds allow only the assistant built into the platform. Others let you bring your own editor and tools.
- Practice with that exact tool. Canva itself recommends practicing with AI tools before the interview. Fluency with a tool you will not be allowed to use does not transfer.
- Plan before you prompt. Spend the first minutes splitting the task into pieces and agreeing on them with the interviewer. Then ask the AI for one piece at a time.
- Read every line it gives you. Say out loud what you are checking. A wrong line you accepted without reading is worse than a slow line you wrote yourself.
- Write tests early. They are the fastest way to show you do not trust generated code blindly.
Where a real-time copilot fits
In a round that allows your own tools, Interview Copilot is one more permitted assistant. It hears the interviewer's follow-up questions and suggests a starting point on screen, and it solves a screenshot of the task. In a round limited to the platform's built-in assistant, use that assistant, because it is the one being evaluated. In a traditional no-AI round, the rules are the opposite, and the invitation will say so.
FAQ
Which companies allow AI in coding interviews?
Canva requires it for backend, machine learning and frontend candidates since June 2025. Meta has been piloting an AI-enabled coding interview. Platforms like CodeSignal offer AI-assisted rounds to any employer that turns them on. Policy is set per company, so the interview invitation is the reliable source.
Can the employer see what I asked the AI?
On platforms with a built-in assistant, usually yes. CodeSignal, for example, gives the employer a full transcript of the candidate's conversation with the AI plus a replay of the session. Treat every prompt as part of your answer.
Do I still need to know algorithms?
Yes, but the emphasis moves. You need enough fundamentals to judge whether the AI's solution is correct and efficient, and to spot the bug when it is not. Canva says it still assesses computer science fundamentals, now through the AI-assisted round.