Skip to content
Hari KharelStudy

Google

Classic algorithms, done properly: one medium to medium-hard problem per 45 minutes, solved optimally with a complexity analysis, then follow-ups that change the constraints. Your code is not run, so you trace it by hand. DP is in scope. No AI. At least one round is now often in person. A hiring committee that never met you decides from written feedback, so a lukewarm packet fails.

The loop

Section titled “The loop”
Stage Format Evidence
Assessments (sometimes) Google Hiring Assessment (“workstyle”), sometimes a coding exercise official
Recruiter 1-2 conversations official
Technical screen 45 min coding reported
Interview panel Video or in person. Typical L5: 2 coding, 1 system design, 1 role-related knowledge or a third coding round, 1 Googleyness & Leadership (45 min each) official panel; reported mix
Hiring committee, then team match The committee reviews the packet and can down-level. Team match can take weeks reported
  • AI: “AI tools are not permitted to be used during your interviews.” official
  • In person (2025 change): “We’ve reintroduced in-person interviews for some roles. In many cases, you’ll complete initial interviews virtually before coming onsite.” official Plan for at least one whiteboard or laptop round.
  • No code execution. You write syntactically correct code and trace test cases by hand. reported
  • Process “typically” takes 6-8 weeks. Interviews are structured: “clear rubrics… no brainteasers.” official

What the coding rounds test

Section titled “What the coding rounds test”

Shape. Interviewers disguise problems and change constraints mid-question; leaked questions get retired. You’re expected to reach the optimal solution and state its complexity. One source puts the bar this way: at L4, correct with some guidance; at L5, the optimal solution found on your own. thin

Topic area How well attested Study first
Arrays and strings, hashing Most common in mock-interview counts reported Hash maps, Two pointers, Sliding window
Trees reported Binary trees, DFS
Graphs: BFS/DFS, topological sort, grids reported BFS, Topological sort, Union-find
Binary search (including “search the answer”) reported Binary search
Dynamic programming and memoisation reported in scope, unlike at Meta 1-D DP, 2-D DP, Memoization
Intervals and scheduling with priority Reported L5 families reported Intervals, Heaps, Sweep line
Tries; string templating; a rate limiter/logger Reported L5 families reported Tries, Parsing, Rate limiting

No company publishes topic frequencies. The order above comes from interviewing.io’s mock-interview counts, which are only a proxy.

System design and behavioral

Section titled “System design and behavioral”
  • System design is expected at L5+, and one source says it is increasingly common at L4. thin Prompts are generic and fundamentals-heavy: a distributed cache, trending hashtags, chat, ticket booking, VM monitoring. Scope the problem before drawing, and commit to choices. reported Study Caching, Distributed thinking and Streaming.
  • Googleyness & Leadership. Use STAR, and quantify: “Accomplished [X], as measured by [Y], by doing [Z].” Expect ambiguity, conflict, influencing others, and leading “even when you weren’t officially the leader.” Questions want “specific examples over ‘I always do X’.” official Sources disagree on how much it weighs. Treat it as a round that can sink you but rarely saves you.

How you’re scored

Section titled “How you’re scored”

The classic attributes are general cognitive ability, role-related knowledge, leadership and “Googleyness”. Today’s pages say only “clear rubrics”, so the current wording is unconfirmed. Interviewers reportedly score on a 7-point scale from Strong No Hire to Strong Hire, and a committee of people who didn’t interview you decides; “five Leaning Hire scores can still result in rejection.” reported

Why people fail reported

  1. Going silent, or skipping the brute-force baseline.
  2. Bugs and missed edge cases in code that is never run.
  3. Not adapting when the interviewer changes a constraint.
  4. Design: drawing before scoping, no commitment, no failure handling.
  5. “I always…” answers in place of a specific story.

Your plan

Section titled “Your plan”

Use the Sandbox’s Google filter. Simulate no execution: write the whole solution, trace two inputs by hand, then press Run, and count the runs you needed. Aim for one run per level. Allow 45 minutes per problem.

2 weeks 4 weeks
Week 1 Calendar Interval Toolkit, Meeting Room Allocation, Minimum Viable Capacity, String Comparisons and Splits, Contract Scheduler, Alien Alphabet Arrays, strings and trees: Pair Targets, Consecutive Runs, Template Filler, Tree Queries, Search Box Autocomplete
Week 2 Search Box Autocomplete, Island Tracker, Rising Runs, Flight Routes, one mock with Repeated Substrings. Two design sessions. Write 6 G&L stories with numbers Graphs and search: Island Tracker, Alien Alphabet, Flight Routes, Minimum Viable Capacity, Calendar Interval Toolkit
Week 3 DP: String Comparisons and Splits, Contract Scheduler, Rising Runs, Sequence Puzzles. Three design sessions
Week 4 Meeting Room Allocation, Repeated Substrings, two mocks on a whiteboard or in a plain doc, G&L mock

Ask your recruiter: Which rounds are in person? Is there a role-related knowledge round or a third coding round? Which languages does the editor support?

Sources

Section titled “Sources”