Yes, but only in one round. Since October 2025, according to Hello Interview, one of the two onsite coding interviews for software engineers and engineering managers up to E7 and M2 is AI enabled, with an assistant in a chat panel beside the editor. The other coding round stays classic, with no AI. Meta has published no public guide to the format.
Which Meta rounds allow AI, and which do not?
The AI enabled round replaces one of the two coding interviews in the onsite loop. Hello Interview describes the loop before the change as two coding rounds, one design round for mid level and above, and one behavioural round. After the change one coding round is AI enabled and the other is the familiar algorithm round, with no assistant.
interviewing.io, in its 2025 survey of interviewers, reported that the AI assisted interview would not replace Meta's algorithmic phone screen and would sit in the onsite instead. No source we found says AI is allowed in the phone screen, the design round or the behavioural round, so assume it is not.
What can the AI assistant do in the round?
It can read the project files and answer in a chat panel. According to Hello Interview it cannot edit files, so every change goes into the editor by your hand, typed or pasted. The guide lists several models from OpenAI, Anthropic, Google and Meta to choose from, and reports that the choice can vary from candidate to candidate.
That design matters for how you practise. If you are used to an agent that edits files and runs commands for you, the round will feel slower. Rehearse with an assistant in a plain chat, where you read its suggestion, decide what to keep and make the edit yourself.
What does Meta grade when AI is allowed?
Hello Interview reports that the AI enabled round is scored on the same four competencies as the classic coding round: problem solving, code quality, verification and communication. It quotes an internal Meta source as saying candidates should use AI but need to show they understand the code.
The same guide reports one candidate getting negative feedback for relying heavily on the assistant, and a recruiter warning that you can be marked down for using it as a crutch. Candidates who checked the AI's output and explained their reasoning were reported to score well. None of this is published Meta policy, so treat it as what candidates and recruiters have said.
How does Meta handle AI in the other rounds?
Meta has gone further than most large employers on policing hidden help. interviewing.io's 2025 survey reported that Meta interviewers must flag suspected cheating across nearly all interview types and levels, and that candidates are asked to share their whole screen and switch off background filters in most interviews.
In the same survey 11% of interviewers at large tech companies said their company uses cheating detection software, and nearly all of them worked at Meta. So the rule is simple: AI where the round is designed for it, nothing hidden anywhere else.
Is the format confirmed by Meta?
Partly. In July 2025 404 Media reported internal Meta posts about the new format, including a call for employees to volunteer as mock candidates in AI enabled interviews. The details of the round, such as the three phases and the chat only assistant, come from prep sites that interviewed candidates. Meta itself has published no guide, so confirm the format, the allowed models and the time limit with your recruiter.
How should you practise for it?
Practise on code you did not write, under a timer, with the assistant in chat only mode. Our invoice totals round starts with debugging existing code, like the first phase of Meta's round, and the rate limit round asks you to extend a small service. Both take 60 minutes and come with checks to run when you finish.
Afterwards, score yourself on the practice rubric, which follows the same competencies: how you broke the problem down, what context you gave the AI, how you verified its output, how you corrected it, and how you explained your choices.