GCA (General Cognitive Ability) is one of the attributes Google evaluates in every interview loop, alongside role-related knowledge, leadership and 'Googleyness'. It is not a separate psychometric test you sit at a desk. It is a way of scoring how you think: whether you can take an ambiguous, open-ended problem, break it into parts, make reasonable assumptions, reason to a defensible answer and explain your reasoning as you go. Because it is assessed inside the coding, analytical and behavioural rounds rather than in isolation, GCA is the attribute candidates most often lose marks on without ever realising it was being measured.
Practise thinking out loud. GCA is scored on process, not on arriving at an answer, so an interviewer needs to hear you clarify the problem, state your assumptions, choose an approach and explain the trade-off before you commit. Structure every answer: clarify, structure, solve, check, then state the answer with its limitations. Practise estimation and open-ended problems, not just questions with clean solutions. And quantify your impact in behavioural answers. Google is explicit about valuing data-driven reasoning. Google does not publish cut scores.
What Is Google's GCA?
Google's hiring is deliberately structured and data-driven: every interviewer scores against the same rubric, and there is no 'personality fit' box to tick, only demonstrated competencies. GCA is the rubric line that captures problem-solving quality. An interviewer scoring GCA is asking: did this candidate understand the problem before attacking it, did they handle the parts they were not given, did they choose an approach for a reason they could articulate, and did they notice their own errors?
The consequence is worth stating plainly. Two candidates can reach the same correct answer and score very differently on GCA, because one of them narrated a structured route and the other produced a memorised solution and could not explain why it worked. Google is hiring for the problems it has not written down yet, and GCA is the proxy for that.
Where GCA Is Assessed in Google's Process
- 1Online application (about 30 minutes). CV and application form, screened for relevant experience and skills.
- 2Online assessment for technical roles (60-90 minutes). A coding challenge on Google's own platform or HackerRank: two to three algorithmic problems, typically medium to hard.
- 3Cognitive ability assessment for some roles (30-45 minutes). A general cognitive ability test used for some business and operations roles, covering numerical reasoning, logical reasoning and problem-solving.
- 4Phone or video screen (45-60 minutes). One or two screens with a Googler: a coding interview for technical roles, a behavioural or analytical interview for business roles.
- 5Onsite interviews (four to five rounds, one day). Technical roles: three to four coding rounds, one system design, one Googleyness. Business roles: a mix of case, analytical and behavioural rounds. GCA is scored across all of them.
Google's loop differs by role, level and region: some roles include a cognitive ability assessment and some do not, and the balance of coding, system design and analytical rounds shifts with seniority. Your recruiter tells you the actual loop composition, and they will usually tell you what each round covers if you ask.
The Four Things Google Actually Scores
- General Cognitive Ability (GCA). How you approach and solve problems, and how clearly you reason. Assessed everywhere, in every round.
- Role-Related Knowledge. Whether you can actually do the job: coding, system design, analytics, product judgement.
- Leadership. Emergent leadership, stepping up to lead when needed and stepping back when someone else should. It is not about job titles.
- Googleyness. Comfort with ambiguity, collaborative instincts, intellectual humility and doing the right thing. It is evaluated in a dedicated round.
The Cognitive Ability Assessment (Some Roles)
For some business and operations roles, Google uses a general cognitive ability assessment of roughly 30 to 45 minutes covering numerical reasoning, logical reasoning and problem-solving. This is a conventional psychometric test rather than an interview, and unlike the GCA attribute it is genuinely practisable: the constructs are the same ones used across graduate assessment worldwide, and speed and accuracy improve reliably with format-specific practice.
| Publisher | Questions | Time | Difficulty | Notes |
|---|---|---|---|---|
| Google coding assessment | 2-3 algorithmic problems | 60-90 minutes | Medium-Hard | Google's own platform or HackerRank. Arrays, strings, hash maps, trees, graphs, dynamic programming |
| Cognitive ability assessment (some roles) | Mixed cognitive items | 30-45 minutes | Medium-Hard | Numerical reasoning, logical reasoning and problem-solving. Practisable in the conventional sense |
| Coding interviews | 3-4 rounds | 45 minutes each | Hard | GCA is scored on how you reason, not only on whether the code runs |
| System design interview | 1 round | 45-60 minutes | Hard | Load balancing, SQL vs NoSQL, caching, CDNs, distributed systems |
| Googleyness / behavioural | 1 round | 45 minutes | Medium | Ambiguity, collaboration, intellectual humility, doing the right thing |
- If your role includes the cognitive assessment, practise it here: free cognitive ability test and logical reasoning.
- For the role-related side of a technical loop: technical assessments.
- Loop composition, roles and levels for the company sit on the Google company page.
Numerical, logical and problem-solving items in one timed set. The free preview gives you 4 questions per test plus 5 AI coach messages, with no payment details required.
Take the free testHow GCA Shows Up in a Coding Interview
In a coding round, role-related knowledge asks whether your solution is correct and efficient. GCA asks something different: how did you get there? The interviewer is scoring the sequence: did you clarify the constraints, did you state an approach before writing, did you weigh at least one alternative, did you reason about time and space complexity, did you test your own code and find your own bug. A candidate who silently types out a memorised optimal solution can score well on knowledge and poorly on GCA.
- Clarify first. Input size, duplicates, negative numbers, guaranteed solution or not. Two clarifying questions cost thirty seconds and change how the whole answer is read.
- State the approach before you code. Name the brute-force solution and its complexity, then say why you are moving to something better.
- Say the trade-off out loud. 'This buys O(n) time at the cost of O(n) space' is a GCA signal in one sentence.
- Test your own code. Walk a small example through it. Finding your own bug scores better than an interviewer finding it.
Worked Example: Reasoning Out Loud
A Google-style coding item. The algorithm itself is well known, which is precisely why it is a GCA test rather than a knowledge test: everyone can produce an answer, so what differentiates candidates is the reasoning around it.
Solve it, and narrate the reasoning the way an interviewer scoring GCA needs to hear it.
Clarify: can the array contain duplicates or negatives, and is a solution guaranteed to exist? Then name the baseline: check every pair with a nested loop, O(n squared) time and O(1) space, correct but wasteful, because for each element you are re-scanning the whole array to find one specific value. State the insight: for each number, the value you need is target minus that number, and 'have I seen a specific value before' is exactly what a hash map answers in constant time. So: iterate once, and for each number check whether target minus number is already in the map; if it is, return the two indices; if not, store the number with its index. Consider the alternatives honestly: sorting with two pointers is O(n log n) time and O(1) space, which is worth mentioning because it wins if memory is the binding constraint, but it destroys the original indices, which the question asks for. Then walk the example: at index 0 you store 2; at index 1 you need 9 - 7 = 2, which is in the map at index 0, so return [0, 1].
A hash map: O(n) time and O(n) space, storing the complement as the key. The correct algorithm is the easy part. What earns the GCA score is everything around it: the clarification, the named baseline, the articulated insight, the explicit space-for-time trade-off, the alternative considered and rejected for a stated reason, and the self-test at the end. Same code, very different interview.
How GCA Shows Up in Analytical and Behavioural Rounds
For business, product and analytics roles, GCA is assessed through open-ended problems: estimation questions, product cases, ambiguous analytical scenarios where you are not given the data you would like. The same structure applies. Clarify what is being asked, break the problem into components, state your assumptions explicitly, work through the arithmetic, then sanity-check the answer against something you know. An estimate you cannot defend is worth less than a rougher estimate whose assumptions you can name and revise.
- Make assumptions visible. 'I will assume roughly a third of users are on mobile. If that is wrong, the answer moves proportionally' is a strong GCA move. Silent assumptions are a weak one.
- Sanity-check your own numbers. Noticing that your estimate implies more users than there are people in the country is a scored moment, not an embarrassment.
- In behavioural rounds, quantify. Google values data-driven decision-making, so 'reduced latency by 40%' beats 'improved performance'. Use the STAR method and put a number in every Result.
- Show intellectual humility. Changing your view when the evidence changes reads as strength here, not as weakness. It is scored under Googleyness and it supports GCA.
How Google Scores You
Interviewers write structured feedback against the four attributes and submit it independently; a hiring committee that did not meet you then reads the packet and decides. Two things follow from that. First, your interviewer is writing evidence, not casting a vote, so anything you do that cannot be written down as evidence is invisible to the people who actually decide. Second, there is no published cut score or pass mark, and there is no single global bar: the standard varies by role, level and team. Any specific score threshold you find quoted online is a guess.
Your Preparation Plan
- 1Week 1: Baseline and fundamentals. If your role has a cognitive assessment, sit one timed and record your accuracy and pace. Technical candidates: return to arrays, strings, hash maps, trees and graphs until the patterns are automatic rather than recalled.
- 2Week 2: Volume with narration. Work medium-difficulty problems, and narrate every one out loud, even alone. Silent practice trains the wrong muscle for a GCA-scored interview.
- 3Week 3: Ambiguity. Deliberately practise problems with missing information: estimation questions, open product cases, under-specified analytics questions. Force yourself to state assumptions before solving.
- 4Week 3: System design (technical roles). Load balancing, SQL versus NoSQL, caching, CDNs, distributed systems. Practise explaining a design and naming its failure modes.
- 5Week 4: Behavioural and Googleyness. Build STAR stories with quantified results, plus specific examples of handling ambiguity, collaborating and doing the right thing when it was inconvenient.
- 6Week 4: Mock loop. Four interviews in one day with a friend, including one where you deliberately get stuck. How you behave when stuck is a GCA signal in its own right.
MakingMoves.ai covers 50+ test categories with 113,000+ practice questions, including cognitive ability, logical reasoning and technical assessments, and the AI coach challenges your reasoning rather than just marking the answer. The free preview gives you 4 questions per test and 5 AI coach messages. Paid plans are GROW at EUR 19,95 per week, PRO at EUR 49,95 per month, and MAX at EUR 79,95 per month.
Common Mistakes
- Solving silently. If the interviewer cannot hear your reasoning, they cannot score it. The most common way strong engineers lose GCA marks.
- Coding before clarifying. Jumping into implementation on an under-specified problem signals that you will do the same thing on the job.
- Reciting a memorised solution. Interviewers ask 'why does that work' precisely to find out whether you know or whether you remember.
- Refusing to make assumptions. In an ambiguous problem, 'I don't have enough information' is not a GCA answer. 'I will assume X, and here is how the answer changes if X is wrong' is.
- Unquantified behavioural stories. Google is explicit about data-driven decision-making. Put a number in every result.
- Freezing when stuck. Being stuck is expected. Saying out loud what you would try next, and why, is scored. Silence is not.
Logical reasoning and problem-solving practice with an AI coach that asks you why, the same question a Google interviewer will.
Explore logical reasoning practiceFrequently Asked Questions
GCA stands for General Cognitive Ability: one of the four attributes Google scores, alongside role-related knowledge, leadership and Googleyness. It measures how you approach problems: whether you clarify, structure, make assumptions explicit, reason to a defensible answer and explain that reasoning. It is assessed inside the coding, analytical and behavioural rounds rather than as a separate test.
The GCA attribute is not a test. It is a scoring dimension applied across your interviews. Separately, Google does use a general cognitive ability assessment of about 30-45 minutes for some business and operations roles, covering numerical reasoning, logical reasoning and problem-solving. That one is a conventional test and it does respond to practice.
Typically five to six: one or two phone or video screens followed by four to five onsite interviews. For software engineering that usually means three to four coding rounds, one system design round and one Googleyness or behavioural round. The loop composition varies by role and level, and your recruiter will tell you.
Googleyness is Google's term for values alignment: comfort with ambiguity, a collaborative instinct, intellectual humility, and doing the right thing. It has a dedicated interview round and it is scored with the same structured rubric as everything else. Prepare specific examples rather than adjectives.
Google does not publish a cut score or a pass mark, and there is no single global bar: hiring committees read structured interviewer feedback against the four attributes, and the standard varies by role, level and team. Any specific threshold you see quoted online is a guess.
Yes, but not by memorising answers. GCA rewards a repeatable process: clarify, structure, assume explicitly, solve, check, and narrate throughout. Practise open-ended and estimation problems out loud, and practise being stuck out loud, because how you behave without a solution in hand is itself a scored signal.
Our assessment tips hub covers timing drills, test-day checklists and preparation plans across all 50+ test categories.
Read the assessment tipsIngmar van Maurik is the founder of MakingMoves.ai and Assessment-Training.com. With 10+ years in psychometric assessment design, he has helped over 1 million professionals prepare for job assessments.