Google's interview structure
Google typically runs 4–5 coding rounds + 1 Googleyness (behavioral) round for SWE roles. Each coding round is 45 minutes with one interviewer. You'll get 1–2 problems. The problems are medium-to-hard by LeetCode difficulty, with a strong emphasis on graphs, trees, DP, and string manipulation. Unlike Amazon, there are no guaranteed easy warmup problems — every round is substantive.
Most common Google problem types
Based on reported interview questions (2023–2025): Graphs (BFS/DFS/Dijkstra) — approximately 30% of problems. String manipulation — approximately 20%. Dynamic Programming — approximately 20%. Trees (especially path and LCA problems) — approximately 15%. Arrays/Two Pointers — approximately 10%. Math/Number Theory — approximately 5%. The graph and DP heavy distribution is distinctly Google compared to Amazon (which skews heavily DP) or Meta (which skews arrays and recursion).
What Google actually evaluates
Google uses a structured rubric with four dimensions: (1) Coding ability — clean, bug-free code written at interview speed. (2) Problem solving — can you approach an unfamiliar problem systematically? (3) Communication — do you explain your thinking before coding? (4) Testing — do you verify your solution with examples and edge cases? Score is on a 1–4 scale per dimension. You need mostly 3s (meets bar) to pass. A single 1 (below bar) usually kills the loop.
Top 20 Google LeetCode problems to study
Word Ladder (BFS, graph) — extremely frequent. Minimum Window Substring — string, sliding window. Meeting Rooms II — intervals + heap. LRU Cache — design + hash map + linked list. Serialize/Deserialize Binary Tree. Course Schedule I and II (topological sort). Longest Substring Without Repeating Characters. Find Median from Data Stream. Trapping Rain Water. Alien Dictionary. Jump Game II. Regular Expression Matching. Word Search II (Trie + backtracking). Decode Ways. Largest Rectangle in Histogram. These appear across many reported Google interviews.
Communication is the differentiator
Two candidates can solve the exact same problem — one passes, one does not. The difference is almost always communication. What Google wants to hear: "Let me think through the brute force first… that is O(n²) time. I notice if I sort this or use a hash map I can reduce the lookup to O(1)… so the overall complexity would be O(n log n)…" Say what you are thinking BEFORE you code it. State the invariant you are maintaining. Flag edge cases aloud before they ask. Call your own bugs during testing.
Preparation timeline
12 weeks is ideal. Week 1–2: Foundations (arrays, strings, hash maps, binary search). Week 3–4: Trees and graphs — spend extra time here for Google. Week 5–6: DP — focus on 1D DP, then 2D, then interval/tree DP. Week 7–8: Hard problems — do at least 20 hard problems under time pressure. Week 9–10: Company-specific practice — filter codingprepguide.com/company to Google problems. Week 11–12: Mock interviews only. Target: 2 mocks per week with a real timer, no pausing, and verbal communication throughout.