AI Interview Index
Which companies let you use AI in their coding interviews, which ban it, and which built the interview around it. Anyone can edit; every change is public and revertible.
Companion to Coding Interviews Should Allow AI on Scaling Knowledge.
- No AI in interviewsBanned in the technical rounds.
- Has AI interviewsAllowed in at least one round.
- AI-native interviewsThe interview is built around it.
33 of 33
| Candidates work in a real codebase and ship a pull request. | San Francisco | ||
| Split by stage rather than company-wide: live interviews are AI-free unless stated, and some take-homes explicitly allow it. | San Francisco | ||
| No coding interview in the traditional sense. Paid contracting on real work, then a work trial, then an offer. | New York | ||
| No coding interview at all, in practice. | New York | ||
| No whiteboard at all: a take-home on a near-real codebase with your own tools, then a paid trial of two to eight weeks at $25/hour. | Remote | ||
| Canva | AI tools are permitted in interviews. Candidates are expected to explain and defend what the tools produced. | Sydney | |
| AI use is expected rather than tolerated. | Sunnyvale | ||
| Self-serve CoderPad assessment as step one, roughly 1-1.5 hours, no AI allowed. | San Francisco | ||
| Rebuilt its question bank after the old ones turned out to be solvable with AI switched on. | Remote | ||
| Keeps AI out deliberately, on the argument that interviews should measure thinking rather than simulate the job. | San Francisco | ||
| Two separate coding rounds on the same day: a 1h "Backend Build" in CoderPad with AI switched off, and a 1h "AI-Assisted Interview". | New York | ||
| No non-AI coding round in the loop. | San Francisco | ||
| Live coding is AI-free by default, but some engineering roles get a designated AI-assisted round. | New York | ||
| Replaced the phone screen, take-home and algorithm onsite with a three-hour on-site build using any AI tools. | New York | ||
| The handbook states plainly that AI use is encouraged in the interview process. | Remote | ||
| Explicitly asks candidates to bring Cursor, Claude, ChatGPT or Copilot, whatever is already in their workflow. | Melbourne | ||
| One round where AI is allowed and a separate round where it is not, deliberately run as two different signals. | New York | ||
| AI use is prohibited by default in technical assessments, per their published careers policy. | Paris | ||
| Pair programming on a realistic problem, with any AI tooling allowed and none required. | London | ||
| One-hour LeetCode-style coding interview after the recruiter call. | Boston | ||
| LeetCode-style screen. No AI assistance. | New York | ||
| A one-hour realistic build on a stubbed codebase, with AI, Google and pairing with the interviewer all encouraged. | South San Francisco | ||
| A 45-minute practical programming problem, explicitly not algorithmic, in any language. | New York | ||
| Splits the signal: a paid SuperDay project with AI allowed, and a separate 45-minute debugging round where it is not. | Remote | ||
| Candidates are handed an agent and a real system and told to make it work, using AI however they see fit. | San Francisco | ||
| Full-stack coding screen plus questions on graph algorithms and HTTP fundamentals. No AI assistance. | New York | ||
| No technical interviews at all. Candidates do a three to five day in-person work trial on real work. | Seattle | ||
| AI use is prohibited by default in technical assessments, per their published candidate policy. | San Francisco | ||
| Replaced coding and algorithm rounds with an AI-native onsite. | San Francisco | ||
| Published per-round rules: AI is banned in some rounds and expected in others. | Geneva | ||
| One hour building a real feature in the Warp product, with Cursor, Claude Code, ChatGPT or any agent fair game. | New York | ||
| Split by design: a whiteboard conversation with no AI, then an implementation round with any AI. | San Diego | ||
| An AI-fluency bar applies at application, screen, skills test and executive interview. | Remote |