The short answer: some coding interviews now let you use AI, or expect you to. Meta’s interviews include a built-in AI assistant, Canva expects candidates to use tools like Copilot, Cursor and Claude, and Google is piloting a round with Gemini. In these rounds the AI can write code quickly, so interviewers score something else: how you clarify the problem, direct the AI, read and check what it produces, and debug it. Many other interviews still ban AI entirely. The rule for every interview: use AI only when the company says you can, and only the tools it provides.
Which companies allow AI in coding interviews
Policies are changing quickly. These are the ones companies have described publicly:
| Company | What it says | Tools |
|---|---|---|
| Meta | “Many of Meta’s interviews now include an AI assistant built into the interview environment. Candidates are expected to use this AI assistant.” | The assistant built into CoderPad, with a choice of Claude, ChatGPT, Gemini or Meta’s models. No outside AI tools. |
| Canva | Since June 2025, it expects backend, machine learning and frontend candidates to use AI tools during technical interviews. | The candidate’s own tools, such as Copilot, Cursor and Claude. |
| A pilot from the second half of 2026 for some junior and mid-level software engineering roles in the US, in the “code comprehension” round, as reported by Business Insider and confirmed by a Google spokesperson. | Gemini, Google’s own model. |
Other companies ask for the opposite. In early 2025, Anthropic’s application form asked candidates not to use AI assistants in their applications, to judge their own communication. If a company hasn’t told you AI is allowed, assume it isn’t, and ask your recruiter if you’re unsure.
What an AI-assisted coding round looks like
The questions change when the candidate has an AI. Canva found that AI assistants “can trivially solve traditional coding interview questions”, so it moved to larger, more ambiguous problems. Its example: instead of implementing Conway’s Game of Life, “Build a control system for managing aircraft takeoffs and landings at a busy airport.”
Expect:
- An existing codebase, not a blank page. Google’s pilot round is about analyzing existing code: finding bugs, improving performance and refining it. Meta tells candidates to be comfortable “reading, debugging, and building on existing code.”
- A bigger, vaguer problem that can’t be solved with one prompt, so you have to break it down.
- The interviewer watching how you work, not only whether the code runs at the end.
What interviewers score
The companies have been specific:
- Google: “AI fluency, including prompt engineering, output validation, and debugging skills”, according to the internal document Business Insider reported on.
- Canva: whether candidates know when and how to use AI, break down ambiguous requirements, make sound technical decisions, and find and fix problems in AI-generated code, and whether the result meets production standards.
- Meta: problem-solving, coding, debugging and collaboration skills.
Canva also described what its best candidates did in its pilot. They asked clarifying questions about the requirements, used AI for well-defined subtasks while keeping control of the overall solution, reviewed and improved the AI’s code, and debugged well when it was wrong. Candidates with little AI experience often struggled, “not because they couldn’t code, but because they lacked the judgment to guide AI effectively.”
How to prepare for an AI-assisted coding interview
- Practice in the same kind of environment. Meta gives candidates a practice session in its interview tool; use it. Otherwise, practice in an editor with an AI chat panel, on a codebase you didn’t write.
- Work on existing code. Take an open-source project, read a module, then fix a bug or add a small feature with AI help. That’s closer to these rounds than solving puzzles from scratch.
- Clarify before you prompt. Say out loud what you understand the requirements to be, the edge cases and what “done” means. Then give the AI a small, well-defined task.
- Read every suggestion before accepting it. Check it against the requirements, run it, and test the edge cases. Explain what you’re checking as you go.
- Keep the design decisions yourself. Let the AI write the parts you’ve specified; you decide the structure, the data model and the trade-offs, and say why.
- Practice debugging AI output. Ask for code, then find what’s wrong with it. It usually has something wrong.
- Keep your fundamentals. Canva says code fluency and technical depth are still requirements. You can’t judge AI-generated code in a language or pattern you don’t understand.
When AI isn’t allowed: the classic coding interview
Most live coding rounds are still traditional. The preparation there hasn’t changed:
- Learn the common patterns, not hundreds of individual problems: two pointers, sliding window, hash maps, BFS and DFS, binary search, heaps, dynamic programming, intervals.
- Talk through your approach first, then code, then test with examples, including edge cases.
- State the time and space complexity of your solution, and whether you could improve it.
- Practice in the real format: a shared editor with no autocomplete, often while talking to someone. Our CoderPad and HackerRank guides cover those platforms.
Preparing without grinding LeetCode
You don’t need hundreds of LeetCode problems. A smaller, deliberate plan works better:
- Work by pattern: solve three to five problems for each pattern above, on any platform, such as HackerRank, Codewars or the company’s own practice test.
- Re-solve from scratch a few days later without looking at your old answer. If you can’t, the pattern isn’t learned yet.
- Practice out loud, with a timer and ideally a peer acting as the interviewer, because explaining while coding is the hard part.
- Know your own code: many interviewers ask you to walk through past projects, so be ready to explain the design choices and trade-offs in something you built.
Take-home assignments are another format, often with more freedom to use AI: see our take-home assignment tips.
Don’t forget the rest of the loop
A software engineering loop is rarely only coding. There’s usually a system design round for mid-level roles and above (see our system design guide) and a behavioral round, where you’ll be asked about projects, conflicts and failures. Use the STAR method for those, and our software engineer interview questions for the full picture.
Frequently asked questions
Can I use AI in a coding interview?
What is an AI-assisted coding interview?
Does Meta allow AI in coding interviews?
Does Google allow AI in coding interviews?
How do I prepare for an AI-assisted coding interview?
Do I still need to study algorithms?
Can I prepare for coding interviews without LeetCode?
Sources
- Meta Careers, Hiring process (questions on AI in technical interviews). Read on 29 September 2026.
- Canva Engineering Blog, Yes, You Can Use AI in Our Interviews. In fact, we insist, Simon Newton, June 11, 2025.
- Entrepreneur, Google Is Testing a New Rule That Could Transform Job Interviews, May 8, 2026, reporting Business Insider.
- Fortune, Anthropic tells job candidates not to use AI, February 4, 2025.
