LeetCode company-tagged questions are lists of problems that LeetCode users report seeing in real interviews at a specific company, grouped by how recently they were reported and, with LeetCode Premium, sortable by frequency. Use them as a final filter, not a curriculum: learn the core patterns first, then spend the last two to three weeks before your loop on the company's recent, high-frequency problems, re-solving them out loud instead of skimming hundreds.
This guide covers where the data comes from, how stale it gets, which time window to pick, a step-by-step way to build a short priority list, and the mistakes that turn company tags into a time sink.
How LeetCode company tags work
LeetCode's own support documentation says the company lists are compiled from a user survey that asks “Have you met this question in a real interview?” and that the lists are kept up to date as survey answers come in. In other words, the tags are crowd-sourced reports from candidates. They are not supplied or verified by the companies themselves.
You reach the lists from a company's page on LeetCode (for example, leetcode.com/company/google) or from the company filter on the problem list. Individual problem pages also have a Companies section that groups tags into recency buckets such as 0–6 months, 6 months–1 year and 1–2 years.
LeetCode's support page lists sorting questions by company as a Premium feature, and the frequency column on company pages is locked for free accounts. The table below reflects the controls on LeetCode's Google company page as archived in September 2025. Labels change over time, so check what your account shows today.
| Control | What it does | How to use it |
|---|---|---|
| Time window | 30 days, 3 months, 6 months, More than 6 months, All | Pick the shortest window that still gives you a few dozen problems |
| Frequency sort | Orders problems from most to least reported in the chosen window (Premium) | Your main ranking signal |
| Role filter | Narrows reports to a role such as backend, frontend or machine learning; the exact role names vary | Use it if your role has enough reports; otherwise leave it on All |
| Difficulty and topic filters | Narrow by Easy, Medium or Hard, or by topic tag | Handy for drilling one weak pattern at one company |
| Updated date | Shows when the list last changed | Check it before you trust a 30-day view; a stale short list is weaker than a fresh 3-month one |
How fresh and reliable the data is
Company tags are the closest thing to a public record of what gets asked, but they have predictable blind spots. Knowing them tells you how much weight to put on any single row.
- They are self-reported. A row means some candidates answered yes to the survey. Nobody checks whether they remembered correctly, whether the real question was a variant, or which round it came from.
- Popular problems collect extra votes. In a public third-party snapshot of the lists taken in July 2026, Two Sum had the top frequency score in the 30-day lists for Google, Amazon and Microsoft, and shared the top score at Meta. The most-solved problems on the site gather the most survey answers, so the very top of a list is partly a popularity signal.
- The long windows are huge. In the same July 2026 snapshot, Google's 30-day list held about 200 problems and its all-time list about 2,300. The All view is an archive, not a study plan.
- Smaller lists are thin. The same snapshot showed 44 problems in Meta's 30-day list and 49 in Microsoft's, against about 200 for Google. At mid-size and smaller companies a short window may hold only a handful of reports, and one candidate can move the ranking.
- Companies reuse questions differently. Candidates commonly report that Meta draws from a recognizable pool of problems, while Google is often said to retire questions once they circulate online. Neither company publishes its question bank, and practice varies by team, level and interviewer, so treat these as tendencies, not rules.
- A title cannot show the follow-ups. Meta's careers-site guide to its engineering interview says coding problems are sized to be solved in roughly 10 to 30 minutes and that the interviewer may add further constraints or requirements once you have a solution. Amazon's SDE prep page asks for syntactically correct code, not pseudocode, with edge cases handled. A tagged problem name tells you none of that.
The right mental model: a company list is a ranked reading list of what candidates remember being asked. It raises your odds of seeing something familiar. It does not tell you what you will get.
When to switch from pattern lists to company lists
Company lists reward people who can already recognize patterns. If you open a 200-problem list before you can tell sliding window from two pointers, every problem is new and little of it transfers. The value is concentrated near the interview: one 2026 review of LeetCode Premium puts the peak at two to four weeks before a specific loop, and for most candidates we suggest the last two to three weeks.
| Time to interview | Main focus | Role of the company list |
|---|---|---|
| 8+ weeks | A pattern list such as Blind 75, Grind 75 or NeetCode 150 | None yet, beyond a glance at which topics show up |
| 4–8 weeks | Finish the pattern list and drill weak topics | Check the 6-month list to see which topics dominate, and weight your drills toward them |
| Last 2–3 weeks | The company's recent high-frequency problems | Main source of new problems, grouped by pattern |
| Final 3–4 days | Re-solves and out-loud practice | Re-solve the top of the list; no new hard problems |
Two exceptions. If your interview is less than two weeks away and your fundamentals are already solid, go straight to the company list but still group it by pattern. If you are interviewing at several companies at once, stay on patterns a little longer and use the overlap between their lists, covered below.
How to prioritize by frequency and recency
The goal is a list of 30 to 50 problems you can actually finish and re-solve. Here is the process we recommend:
- Pick the window by company size. For the largest employers, start with 30 days or 3 months. For mid-size companies, use 3 to 6 months. For smaller companies, 6 months is often the shortest window with enough reports to rank.
- Apply the role filter if it has data. A frontend or machine learning loop can lean differently from a backend one. If filtering leaves only a few problems, turn it off.
- Sort by frequency and take the top 40 or so. Stop there. The bottom of a long list is mostly single reports.
- Cross-check against a longer window. A problem that ranks high in both the 30-day and the 6-month view is a steadier signal than one that only spiked this month.
- Discount the famous warm-ups. If an easy classic tops the list, solve it once to confirm you can, then move on. Its rank reflects popularity as much as interviews.
- Group what is left by pattern. Solve all the interval problems together, then all the graph problems, and so on. Batching teaches a pattern faster than random order, and our guide to the 15 core LeetCode patterns lists the trigger for each one.
- Order by difficulty inside each group. Mediums first. Add a hard only if it ranks high in more than one window.
A quick way to score each problem on your list:
| Signal | Priority |
|---|---|
| High frequency in both the short window and the 6-month window | Do first and re-solve twice |
| High in the short window only | Do next; confirm it is not just a famous easy problem |
| Appears only in the 6-month window | Do if time allows |
| Appears only under More than 6 months or All | Skip unless it fills a pattern gap |
Interviewing at more than one company
Pull the top of each company's recent list and look for overlap. Problems that appear on two or three lists go first, and the topics they share tell you where to spend drill time. Then add a small company-specific set in the final week before each loop.
Turning the list into a two-week routine
A tagged list only helps if you can produce the solution under interview conditions. A daily block that works for most people:
- Two new problems from the list, timed. Give each about 30 minutes. If you are still stuck after that, read the approach, close it, and re-solve from a blank editor the next day.
- Two re-solves from earlier days, with no notes. A problem counts as done when you can rewrite it cold, not when it shows a green check.
- One follow-up drill. Take one problem and change a constraint: the input no longer fits in memory, the array arrives as a stream, or you need the k best answers instead of one. That is the kind of added constraint Meta's guide warns about.
- Some practice in a plain editor. Meta's guide suggests a simple text editor because its interview environment has no syntax highlighting or auto-completion, and Amazon asks for syntactically correct code.
- Talk the whole time. State the brute force, the better approach and the Big-O before you type.
Every few days, rehearse a full round out loud. A free AI mock interview in CoPilot Interview works for this: paste the job description and your résumé, set the mix to coding plus whatever else the loop includes, answer out loud, and each answer gets a score with what worked and what to improve.
Pitfalls that waste company tags
- Memorizing code instead of the idea. A memorized solution falls apart the moment the interviewer changes a constraint. You should be able to explain why the approach works and what it costs.
- Studying the All window. Thousands of problems, many reported years ago. You will never finish it, and much of it no longer reflects the current loop.
- Starting too early. Company lists before pattern fluency produce slow, scattered practice. The lists are also refreshed regularly, so problems you studied two months out may have dropped off by interview day.
- Buying Premium months ahead. If company tags are the only reason you want it, a single month bought once the interview is scheduled usually covers it. Our breakdown of whether LeetCode Premium is worth it covers what else it unlocks; check current pricing on LeetCode's subscribe page before you pay.
- Chasing the hard tail. One report of an obscure hard problem is noise. Spend that hour re-solving a high-frequency medium.
- Forgetting the rest of the loop. A strong coding round rarely carries an onsite alone. Leave real time for behavioral and system design rounds.
If you do not have Premium
You can still get most of the benefit:
- Pair a pattern list with a company guide. Our guides to Google coding interview questions, Meta coding interview questions and Amazon coding interview questions cover the topics and the bar for each loop.
- Read the company's own prep material. Meta's careers site and Amazon's SDE interview prep page both describe what the coding rounds look like and what interviewers evaluate.
- Read recent interview reports critically. Forum posts are the same kind of self-reported data as company tags, with even less structure. Weigh topics that recur across many posts, not single stories.
- Check the date on any free copy. Public GitHub snapshots of the company lists exist, but each one is frozen on the day it was taken. A snapshot from last year behaves more like an all-time list than a 30-day one.
Rehearse your company list out loud
Paste the job description and your résumé, pick the coding mix and answer out loud. Every answer is scored with what worked and what to improve, and the session report lists likely next-round questions and practice drills. The free plan gives you 15 minutes a day with no credit card.
Try a free AI mock interviewFAQ
Are LeetCode company-tagged questions accurate?
They are self-reported by LeetCode users through an in-site survey, so they show what candidates say they were asked, not an official list from the company. They are useful for spotting recurring problems and topics, but expect variants and follow-up questions.
Do you need LeetCode Premium to see company tags?
LeetCode lists sorting questions by company as a Premium feature, and the frequency sort on company pages is locked for free accounts. If you only need it for one loop, a single month bought close to the interview is usually enough.
Which time window should I use: 30 days, 3 months or 6 months?
For the largest companies, start with 30 days or 3 months. For smaller companies with fewer reports, use 6 months so the list is not too thin. Avoid the All view, which mixes in reports that are years old.
When should I start practicing company-tagged questions?
In the last two to three weeks before the interview, once you can already recognize the core patterns. Earlier than that, a pattern list gives you more transferable practice per hour.
How many company-tagged problems should I solve?
Around 30 to 50 from the top of your chosen window, each solved and then re-solved from scratch, is a better use of two to three weeks than skimming 150. Problems you can rewrite cold are worth more than problems you have seen once.
Will I get the exact same question in my interview?
Sometimes, but do not count on it. Candidates commonly report that some companies repeat questions more than others, and interviewers often change constraints or add follow-ups, so practice explaining the pattern rather than memorizing code.