Google brings AI into interviews

Google is preparing to let some software-engineering candidates use artificial-intelligence tools during parts of the interview process, according to reports published on May 7 and May 8 and comments confirmed by the company. The pilot will let candidates in certain coding rounds use Google’s Gemini assistant rather than treating AI as an off-limits aid. (businessinsider.com) The change is aimed at select U.S. teams hiring for junior and mid-level roles later in 2026. Google said the new format is intended to better match how its engineers already work. (b17news.com) The move comes as Google has publicly described AI as a routine part of its own software-development workflow. (briefs.co) Sundar Pichai said on April 22 that 75% of all new code at Google is now AI-generated and then approved by engineers. Google Cloud’s 2025 DORA report, published in September 2025, said 90% of software-development professionals surveyed use AI in core workflows and more than 80% reported productivity gains. ### Which interview rounds are changing first? The new pilot centers on a “code comprehension” round in which candidates read, debug and improve an existing codebase with Gemini available as an approved assistant, according to reports citing internal guidance and company confirmation. (businessinsider.com) Interviewers are expected to judge not only the final answer but also how candidates use the tool during the exercise. Select U.S. teams hiring for early- and mid-career roles are expected to go first, with possible expansion to more groups and regions if the pilot works, according to the reports. (blog.google) Briefs, citing Brian Ong, said the first organizations to test the format include Google Cloud and the company’s Platforms and Devices unit. ### What will interviewers be looking for besides code? Internal guidance described in multiple reports says interviewers will evaluate “AI fluency,” including prompt engineering, output validation and debugging skills. The format has been described as “human-led, AI-assisted,” a phrase repeated across accounts of the pilot. (careers.northeastern.edu) That means candidates are still expected to explain what the model produced and why they accepted or rejected it. Reports on the new process say the interview is designed to expose whether a candidate can inspect AI output, spot mistakes and reason through trade-offs instead of simply pasting in generated code. (briefs.co) ### Why is Google changing interviews now? On April 22, Pichai said Google had moved from 50% of new code being AI-generated last fall to 75% in spring 2026, with engineers still reviewing and approving that work. That public benchmark gives context for why the company would want interviews to test AI-assisted coding rather than unaided puzzle solving alone. (careers.northeastern.edu) Google’s own research has pointed in the same direction. The 2025 DORA report said 65% of surveyed software professionals rely on AI at least moderately for development work, while 59% said AI had a positive effect on code quality, even as trust in outputs remained uneven. (careers.northeastern.edu) ### Does this mean candidates can let Gemini do the work? Reports about the pilot say no. Candidates may use Gemini as an approved assistant, but the interview still probes whether they understand the code, the bug, the fix and the limits of the model’s answer. Accounts of the process describe interviewers pushing candidates to explain their reasoning, which makes AI-only answers risky if the candidate cannot defend them. (blog.google) That approach mirrors the tension visible in Google’s own research. The DORA report said developers widely use AI but do not uniformly trust it, suggesting that checking outputs remains part of the job rather than something AI removes. (blog.google) ### Are other parts of the hiring loop changing too? The 2026 interview overhaul described in the reports goes beyond the AI-assisted coding round. One change would add a more technical design discussion to the long-running “Googleyness and Leadership” interview, and another would replace one traditional technical interview for junior candidates with a more open-ended engineering session. (b17news.com) Those reported changes would move more of the loop toward code review, debugging and engineering judgment. Exponent’s guide, republished by Northeastern University, said Google had already been moving away from pure data-structures-and-algorithms screens toward code comprehension and ambiguous problem-solving. (blog.google) ### What happens next in the rollout? Later in 2026, Google plans to begin the pilot with select U.S. teams and then consider a broader expansion if the format performs as intended, according to the company comments cited in multiple reports. During the pilot phase, candidates will use Gemini as the approved assistant. (careers.northeastern.edu) The first named groups are Google Cloud and Platforms and Devices, according to Briefs’ account of Ong’s comments. Any wider rollout would come after those teams test the format in live hiring. (briefs.co) (msn.com) (careers.northeastern.edu)