Google's In-Person Interview Shift: What It Means for AI-As
Google's interview process is moving in two directions at once, and candidates need to understand the tension before they walk into either room. On one side, the company has been rolling out AI-assisted formats like the Code Comprehension round, where candidates use Gemini to navigate unfamiliar codebases and demonstrate AI fluency . On the other, CEO Sundar Pichai confirmed that Google will introduce at least one in-person interview round for engineering roles, explicitly citing concerns that candidates are misusing AI tools during online coding assessments . The result is a hybrid process that tests whether you can work with AI in one round and whether you can prove your fundamentals without it in another. Understanding why both exist—and how to prepare for each—is now essential for anyone targeting Google engineering roles.
The driver behind the in-person shift is straightforward: virtual technical interviews have become unreliable. Pichai first raised the idea during a February town hall, where employees directly asked leadership to bring back onsite interviews, citing email threads about compromised assessments . On the Lex Fridman podcast in June, he made it official: "We are making sure we'll introduce at least one round of in-person interviews for people, just to make sure the fundamentals are there" . The phrasing matters. The in-person round is not designed to test whether you can use AI well—that is what the Code Comprehension round evaluates. It is designed to verify that the skills underneath the AI assistance are real. Some hiring managers have reported that over half of candidates in virtual technical interviews may be using unauthorized AI tools, making it difficult for interviewers to distinguish genuine problem-solving from off-camera assistance . The in-person requirement is a filter against that specific failure mode.
This creates an apparent contradiction that candidates must navigate carefully. Google says AI fluency is now a requirement for all software engineering jobs, and its internal documents describe the Code Comprehension format as "human-led, AI-assisted" . Yet the same company is adding face-to-face rounds partly because AI use in interviews has become a problem. The resolution lies in the distinction between using AI as a tool and relying on it as a substitute. A LeetCode discussion from November 2026 captured this precisely: a candidate for a similar AI-assisted round at another company was rejected because they started prompting before understanding the codebase, and the AI's suggested fix didn't actually solve the problem. What was graded, according to that account, was "how you drive the AI, not how much code it writes" . The in-person round exists to confirm that the person driving the AI actually understands what the AI is producing.
For candidates, the practical implication is that preparation can no longer focus on a single skill set. The AI-assisted rounds reward code reading, precise prompting, and critical validation of Gemini's output . The in-person rounds reward exactly what those AI-assisted rounds cannot fully verify: real-time reasoning without a side panel, whiteboard problem-solving without a prompt window, and the ability to explain your thinking when no tool is generating suggestions for you. Pichai framed the hybrid approach as beneficial for both sides: "Given we all work hybrid, I think it's worth thinking about some fraction of the interviews being in person. I think it'll help both the candidates understand Google's culture and I think it's good for both sides" . From a candidate's perspective, the in-person round is also an opportunity to demonstrate the communication and collaboration skills that AI tools cannot simulate.
The broader industry context reinforces that this is not a temporary experiment. Amazon now requires candidates to acknowledge they will not use unauthorized tools during interviews, Anthropic explicitly prohibits AI use in applications, and Cisco, McKinsey, and Deloitte have all added or restored face-to-face components . Google's move is part of a wider correction: companies that spent years optimizing for remote convenience are now optimizing for assessment integrity. For candidates, that means the era of preparing exclusively for one format is over. The most effective preparation strategy is to build both capabilities deliberately—practice reading and debugging code with AI as an assistant, and practice solving problems without it, narrating your reasoning as you would in a room with an interviewer. The candidates who succeed in Google's hybrid process will be the ones who can demonstrate that their skills are genuinely theirs, whether Gemini is in the room or not.sisted Rounds
