Google is testing a different kind of software engineering interview: in selected coding assessments, candidates can use an AI assistant while interviewers evaluate how well they direct, verify, and debug its work. The trials have spread beyond their initial small-group tests, but Google still describes the formats as pilots.
4
6
13
What changes in the coding assessment
The pilot centers on a “code comprehension” round. Instead of assessing only whether a candidate can write code from scratch, the format uses an existing codebase and allows an approved AI assistant. Business Insider’s May report described the initial test as aimed at junior- to mid-level candidates on selected U.S. teams.
9
13
The reported evaluation focuses on practical AI fluency: writing useful prompts, checking the assistant’s output, and debugging problems. In other words, using AI is not enough; candidates are expected to assess and improve what it produces.
9
11
The trials are expanding, but remain pilots
Business Insider reported in October that the interview changes had reached more teams. The same reporting says Google still considers the new formats pilots, so candidates should not assume that AI is allowed in every Google interview or coding round.
4
6
The changes also reach interviewers’ work: internal-document reporting says Google is permitting interviewers to use Gemini to prepare candidate feedback. The reporting supports this specific use; it does not establish every AI-assisted step sometimes attributed to the hiring process.
6
What “AI fluency” means—and what the evidence does not show
Business Insider reported that many engineers are expected to show some degree of AI fluency. That supports a broader shift in expectations, but it does not establish that AI fluency is mandatory for every engineering job or that most roles above entry level require AI-product integration skills.
6
Google’s reported workplace use of AI helps explain the relevance of the interview changes: in April, the company said AI generated three-quarters of its new code, with human engineers reviewing it. That figure describes code generation and review—not autonomous engineering or automated hiring decisions.
14
Details such as AI-assisted interview transcription and note-taking, rules against building standalone hiring agents, and the exact mix of traditional and AI-assisted rounds are not established by the reporting cited here. The clearest supported takeaway is narrower: Google is testing whether candidates can work with AI while still exercising engineering judgment, and the tests have not yet become a universal hiring rule.
4
6
13