Google Interview Questions
& Process for Software Engineers

Nobody who interviews you decides whether you get hired. They write you up and assign ratings, and then four or five engineers who never met you read those write-ups and make the call. Which means the thing being judged is not really the hour you spent solving problems. It is the paragraph your interviewer types afterwards, from memory, between two other meetings. Seven stages, six to twelve weeks end to end, and team matching alone can stall for months when hiring is frozen.

Roughly 90% of resumes are filtered out before a recruiter opens them. Phone screens happen in a shared Google Doc, which is about as pleasant as it sounds. The onsite is four to five 45-minute rounds, and one of them is a behavioral interview called G&L, for Googleyness and Leadership. Candidates consistently underrate it, and it decides as much as any coding round does. Google describes the overall flow on its own how we hire page.

Down-leveling is common enough that you should plan for it. The HC sets your level and your interviewers only weigh in, which is why a team's Statement of Support does not settle the question. One candidate with eight years of experience went from L5 to L4 with one in hand. If you expect to sit on the border, targeting the lower level and promoting internally is usually the better trade, because promotion from inside is far easier than appealing a level after the fact.

Scope is the thing that moves you. Describing IC work when the level calls for technical leadership will pull you down, and so will a system design round where you never reach the trade-offs. Quantified impact pulls the other way, especially stories where you clearly operated above your title. Treat the bands below as rough, since real offers cluster near the middle of each one.

Total compensation per Levels.fyi (US, October 2026); self-reported and updated continuously.

The editor is a Google Doc. That sounds like a gimmick right up until you are twenty minutes in, hunting a bracket that nothing will highlight for you, while someone watches you do it. Each round runs 45 minutes. Most candidates spend forty of them writing and then run out of time before testing, which is exactly backwards, because the test pass is part of what gets graded.

Interviewers grade on four axes: Coding (does it work, is it clean?), Communication (can you think out loud?), Problem-solving (do you decompose well?), and Algorithm knowledge (do you reach for the right pattern?). All four land in the written feedback the Hiring Committee reads, and that is where the damage usually hides. Three strong scores with nothing written about the fourth reads to the committee as a gap you were hoping nobody would mention.

“Excessive hinting is a negative signal. The HC distinguishes between ‘needed one nudge’ and ‘couldn't progress without hand-holding.’”

The ratings run Strong Hire, Hire, Lean Hire, Lean No Hire, No Hire, and four or five engineers at L6+ read the full written feedback behind them. Detail beats the label more often than people expect: a Lean Hire written with real enthusiasm outweighs a Hire backed by two generic sentences. Roughly 60% of candidates who reach this stage are rejected anyway.

A packet of all Hire and Strong Hire is close to automatic. Three positives with one Lean No Hire is the genuine borderline, two Lean No Hires is almost always fatal, and any outright No Hire is very hard to survive. A behavioral flag from G&L overrides all of it. Candidates with perfect coding scores have been turned down over a single culture concern.

Since the committee never meets you, the written feedback is the interview. An interviewer who liked you and then wrote two generic sentences does more damage than one who leaned negative and wrote three specific paragraphs. Which means the draw matters more than it should: a junior interviewer who writes carefully will often do more for you than a senior one who phones it in.

An offer has four parts: base, a 15% annual bonus, sign-on, and GSUs. Equity is the line with the most give in it, and Google can raise a grant 30-50% without anyone needing extra approvals. Vesting is front-loaded at 33/33/22/12 on most grants with no cliff since 2021, though ask your recruiter for the exact schedule rather than assuming yours matches.

A competing offer from Meta or Apple is the strongest lever you will have, though the top-of-band question is worth asking without one. Plenty of people accept the first number and leave real money behind. Time the conversation for after team match, while the offer is still being written.

Recruiters expect a counter, and most carry a 10–15% buffer they can release without escalating. The common mistake is aiming that counter at base, where the bands are narrow and the resistance is real. Equity and sign-on are where the movement actually is. The “top of band” question matters most at L5 and L6, where the spread between the low and high end can run past $100K a year.