Canva's AI-Assisted Coding Interview: What to Expect

Canva is one of the few companies that has described its AI-assisted coding interview in its own words. It does not just allow AI in this round. It expects you to use it. This guide sticks to what Canva's engineers have published, points out what they have not, and gives you a way to prepare. For how other companies set their AI rules, see our AI coding interview overview, and for the format in general, what an AI coding round is.

AICodingRound is not affiliated with Canva, and we have no copy of Canva's interview questions.

Practice the judgment half on AICodingRound. Canva scores whether you own your code and can critically review what an AI wrote. Our problems train exactly that, in real multi-file apps. A good start is Review the AI's reminders PR (medium, about 30 min). The beta is free and places are limited: get a place.

What Canva has said on the record

Two posts on Canva's engineering blog are the primary sources. Everything in this table comes from them.

The first post explains the reason. Canva saw candidates using AI anyway, sometimes "covertly through tools specifically designed to avoid detection," and chose transparency over detection. In internal tests, AI tools solved its old fundamentals questions and "produced correct, well-documented solutions in seconds." News coverage at the time added that Newton is Canva's head of platforms and that some Canva engineers worried the change would replace fundamentals with "vibe-coding sessions" (The Register). Canva's answer was that code fluency and technical depth are still required, just tested in a different setting.

What the round asks you to do

The example Canva gives: instead of a classic puzzle like Conway's Game of Life, candidates might be asked to "Build a control system for managing aircraft takeoffs and landings at a busy airport." The problems are meant to be more complex, more ambiguous and closer to real product work.

You use your own preferred AI tools. The second post, written by an interviewer, splits a strong interview into three phases: planning, development and discussion. In plain terms:

• Planning. Explore the problem space and clarify the requirements.
• Development. Use AI "for code generation, research, and routine implementation tasks," then "review and understand everything that gets generated." Aim for code that could be submitted as a pull request for peer review.
• Discussion. "Be prepared to explain and justify your technical choices," including what you would improve with more time and how the solution would scale.

What interviewers evaluate

The interviewer's post names three areas, each with a guiding question (Canva Engineering):

1. Engineering problem-solving. "How do you approach ambiguous, realistic problems?"
2. Technical depth and ownership. "Do you understand and own the code in your solution?"
3. AI collaboration effectiveness. "Can you critically evaluate AI outputs and iterate toward better solutions?"

The first post describes what the most successful pilot candidates did: they "Asked thoughtful clarifying questions about product requirements," "Used AI strategically for well-defined subtasks while maintaining control of the overall solution," and "Critically reviewed and improved AI-generated code." Canva also notes that "candidates with minimal AI experience often struggled," and says the problem was judgment about guiding the AI and spotting weak suggestions, not coding ability.

The interviewer's post also names three approaches that "don't shine":

• AI Showcase: spending the interview showing off the tool instead of solving the problem.
• Feature Marathon: "We'd rather see thoughtful problem-solving and quality code than a rushed feature list."
• Hands-Off: treating AI as "a complete solution generator rather than a development accelerator."

One more line is worth remembering: "If your interviewer asks you a question, tell us what you think before getting help from AI."

What is not public

Canva's posts do not give the length of the round, the languages offered, whether you work on your own machine or a shared environment, or whether tests are expected. The first post says Canva expects machine learning candidates to use AI tools too, but it names backend and frontend roles as the ones where this round replaces the fundamentals screen. Candidate reviews on job sites describe the round, but we could not open them to check, so we have left their details out. Ask your recruiter about anything on this list.

How to prepare

Set up before the day. The interviewer's post says to have "your preferred AI-integrated development environment ready" and to set up a basic project structure in your language, so you do not spend interview time on setup. Make sure you know your tool's limits, not just its strengths.

Practice the planning out loud. Take a vague prompt (a parking garage, a library loan system, a delivery tracker) and spend five minutes on questions before writing code. What are the entities? What can go wrong? What is out of scope?

Give the AI small, clear jobs. Ask for one module or one function with stated inputs and constraints, not the whole system. Then read what came back as if a teammate wrote it.

Review like you will be asked about it. You will be. For each piece the AI wrote, be able to say what it does, one way it could fail, and why you kept it.

Leave time for the discussion. Stop building with a few minutes left. Say what you would test, what you would change, and what breaks first at ten times the load.

Practice for Canva's round on AICodingRound

Canva's three areas are problem-solving, owning your code, and evaluating AI output. AICodingRound trains all three on real code: you clarify a ticket, build or fix inside an existing multi-file app, and review what an AI wrote. Your own Claude or Codex works as a coach that explains but does not write the code, so every line you keep is one you can defend. One note: Canva's round is an open-ended build where the AI may write code, so also rehearse a from-scratch build with your own tools.

• Book a room every week (medium, Python, about 45 min). A feature with real edge cases to clarify.
• Save a cart for later (PRD) (medium, TypeScript, about 50 min). Build from a product spec.
• Refund reasons you can correct later, and a webhook when they change (medium, Python, about 45 min). A feature that touches an existing contract.
• Review the AI's reminders PR (medium, Python, about 30 min). Find what an AI got wrong in a feature.
• Review the AI's retry PR (medium, Python, about 10 min). A short review drill.

Every problem ends with a short debrief about your own change, which is good practice for Canva's discussion phase, and a report with a level per skill (Strong, Solid or Developing), including how you used AI. Our guides on reviewing AI-written PRs and AI fluency interviews go deeper on the review and judgment parts.

Start with one feature and one review this week. The beta is free and places are limited: get a place.