How Google Is Changing Its Interview Process for the AI Era

For years, Google's engineering interview was synonymous with whiteboards, algorithms, and a candidate explaining every step aloud. The pandemic pushed much of that ritual onto video calls. Now AI is pushing part of it back into the room.

In a June 2025 conversation with Lex Fridman, Google CEO Sundar Pichai said the company would introduce at least one in-person interview round “just to make sure the fundamentals are there.” That is the most direct public confirmation of the change. It is not a declaration that every interview will be on-site, nor that Google has abandoned virtual screening. It is a move toward a hybrid process in which at least one stage gives the company a higher-confidence view of how a candidate actually thinks.

The important story is not simply that office interviews are back. It is that Google is trying to separate two abilities that AI has made harder to distinguish remotely: understanding the work and operating the tools.

Why Google is bringing a round back in person

Remote interviews are convenient, widen the talent pool, and reduce travel. They also make it difficult to know what is happening outside the camera frame. A candidate can now receive a polished explanation or a complete solution from an AI assistant in seconds. More elaborate systems can listen to an interview and surface suggestions in real time.

That weakens the signal from a traditional coding question. If the exercise is meant to reveal decomposition, trade-offs, and debugging ability, a correct final answer is not enough. The interviewer needs confidence that the reasoning belongs to the candidate.

This problem is larger than Google. Axios reported in August 2025 that Google, Cisco, and McKinsey were among the companies bringing face-to-face meetings back into parts of hiring as employers responded to AI-assisted cheating and candidate fraud. Google's answer is not to move the entire pipeline backward. It is to add a verification point.

Fundamentals still matter—but the definition is changing

Pichai's answer had a second half that is easy to miss. He said the ability to use AI tools to generate better code is also an important skill and an asset.

That creates a two-part bar for modern engineers:

1. Can you reason about a problem without outsourcing the understanding?
2. Can you use AI to improve the speed or quality of your work without trusting it blindly?

These are not contradictory requirements. An engineer needs enough knowledge to notice when generated code mishandles an edge case, introduces a security problem, or solves the wrong problem elegantly. AI fluency is most useful when it sits on top of technical judgment.

The in-person round protects the first signal. The day-to-day job increasingly rewards the second.

What probably has not changed

Candidates should be careful with claims that Google's entire interview loop has been reinvented. The publicly confirmed change is at least one in-person round; implementation can vary by role, level, team, and location.

Google's own candidate resources still describe virtual interviews and technical collaboration through Google Meet and Google Drawings. Its re:Work guidance continues to emphasize structured interviews: consistent questions, clear scoring rubrics, and follow-up prompts designed to expose a candidate's thought process. Those principles remain useful whether the conversation happens through a screen or across a table.

So the safest interpretation is evolution, not replacement. Recruiter conversations, assessments, technical screens, behavioral evaluation, and role-specific rounds may still appear. What changes is that a fully remote performance is less likely to be the only evidence used for a hiring decision.

How candidates should prepare now

The wrong response is to grind more memorized answers. An in-person verification round makes rehearsed fluency less valuable than adaptable reasoning.

Practice explaining the decision, not just writing the code

Talk through the constraints you notice, the alternatives you reject, and the tests you would run. Interviewers can evaluate a partially completed solution if the thinking is clear. Silent speed provides much less evidence.

Rebuild the fundamentals you usually delegate

Be able to reason about data structures, complexity, APIs, concurrency, databases, and system behavior without immediately asking an assistant. The goal is not to reject AI. It is to retain the mental model required to check its work.

Use AI as a reviewer during preparation

Ask an assistant to challenge assumptions, generate adversarial test cases, or critique a design. Then verify its claims. This trains the skill Pichai described: producing better work with the tool while keeping responsibility for the result.

Prepare for an actual room

Practice with a whiteboard or a plain editor, maintain eye contact while thinking, and become comfortable asking clarifying questions aloud. In-person interviews add ordinary human factors—travel, nerves, body language, and unfamiliar equipment—that do not appear in a LeetCode score.

The bigger shift in technical hiring

AI has not made engineering interviews irrelevant. It has made weak interview signals easier to fake.

Google's response points toward a broader model for hiring: verify core ability in a controlled setting, then evaluate how effectively a candidate uses modern tools. A company that tests only unaided coding risks selecting for a work style that no longer reflects the job. A company that permits unrestricted assistance without checking understanding risks hiring the tool's output instead of the person.

The durable skill sits between those extremes. Strong candidates can think independently, collaborate clearly, and use AI without surrendering judgment. Google's changing interview process is an early attempt to measure all three.

Sources

• Lex Fridman: Sundar Pichai interview transcript
• Axios: Companies embrace in-person interviews to dodge the chatbots
• Google candidate guide for virtual interviews
• Google candidate guide for technical virtual interviews
• Google re:Work: A guide to structured interviewing