CCAO-F Exam Domains: Complete Breakdown of All 7 Domains [2026]

The Claude Certified Associate - Foundations (CCAO-F) exam distributes 60 scored questions across seven domains, and each domain carries a published weight that determines how much of your final score it controls. Before you build a study schedule, you should know those weights cold, because they answer the most practical question in exam preparation: where do your points live?

This guide breaks down every CCAO-F domain: the exact weights, the approximate question counts they produce on a 60-question exam, the topics inside each domain, and the business scenarios the exam wraps them in. The exam runs 120 minutes and requires a scaled score of 720 out of 1000 to pass (scaled scoring converts your raw correct-answer count onto a common scale so results stay comparable across exam forms). It mixes single-answer multiple choice with multiple-response items that ask you to Select TWO or Select THREE, roughly a quarter of the total.

CCAO-F certifies professionals who use Claude through the claude.ai product as a working tool: people in marketing, operations, project management, education, and communications. Every domain is tested through everyday business situations such as drafting client emails and preparing reports, and no domain requires coding. If the certification itself is new to you, start with the complete CCAO-F certification guide or the shorter what is CCAO-F overview, then return here for the domain-level detail.

What this guide covers:

• The complete weight breakdown for all 7 domains with approximate question counts
• Every topic each domain tests, taken directly from the exam blueprint
• Concrete business scenarios showing the judgment behind the questions
• A weight-based study prioritization strategy with links to a full preparation plan

Domain Weight Overview

The weights below come straight from the exam blueprint. To convert a weight into a question count, multiply it by 60 and round: Output Evaluation and Validation at 21% works out to roughly 13 questions, while Troubleshooting and Optimization at 10% still contributes about 6. Exact counts can shift by a question between exam forms, but the proportions hold.

The Numbers That Matter

One domain towers over the table. Output Evaluation and Validation (21%) contributes roughly 13 of your 60 questions, twice the weight of the lightest domain, and the message is worth absorbing before anything else: Anthropic considers the ability to judge, verify, and refine what Claude produces the defining skill of a certified associate.

The middle band covers daily practice. Workflow Integration and Solution Design (16%), Governance, Risk, and Responsible Use (15%), and Prompting and Task Execution (14%) together control 45% of the exam, about 27 questions, and they test how you bring Claude into real work. The remaining product-knowledge band (Product and Model Selection, Configuration and Knowledge Management, and Troubleshooting and Optimization, 34% combined) rewards hands-on familiarity with claude.ai features, and it studies fastest because every topic can be practiced inside the product.

Domain 1: Output Evaluation and Validation (21%)

Approximate question count: ~13 of 60

Domain 1 is the heaviest domain because everything else depends on it. Claude produces fluent, confident text whether or not that text is correct, and the blueprint expects a certified associate to treat every output as a draft to be verified rather than a finished fact. The central concept is the hallucination: a statement the model presents confidently even though it invented the detail instead of drawing it from a reliable source. Hallucinations are dangerous in business settings precisely because they read smoothly.

The tested judgment shows best in a concrete case. Suppose Claude drafts a follow-up email to a client after a project delay: it apologizes gracefully, proposes a revised timeline, and references a phone call in which the client supposedly approved the new scope. The tone is excellent, and the phone call never happened. The exam checks whether you separate judgment content (tone, structure, phrasing), which you edit like any writing, from verifiable claims (dates, figures, names, prices, commitments), which you trace to a source before anything leaves your outbox.

Deciding when human review is required is the most tested call in the domain, and the principle is proportionality. A brainstorm of campaign taglines can ship after a light skim, while anything legal, financial, medical, contractual, or promised to an external party needs a qualified person to sign off. Bias identification applies the same discipline to people: summaries that describe two equally strong employees in noticeably different language deserve the same scrutiny as a suspect statistic.

Output formats round out the domain. Artifacts (the claude.ai feature that opens a generated document or table in its own panel, where you can edit, iterate, and export it) fit deliverables that will live beyond the conversation. An inline response fits a quick answer you read once, and a structured table fits information someone will compare or sort.

What to study: the claude.ai Help Center material on accuracy and Claude's limitations, then fact-check three real Claude outputs from your own work. The habit that exercise builds is the exact habit this domain measures.

Domain 2: Workflow Integration and Solution Design (16%)

Approximate question count: ~10 of 60

Domain 2 tests whether you can look at a business process and find the places where Claude genuinely helps. The blueprint starts with requirements analysis: defining the goal, the available inputs, the quality bar, and the deadline before opening a chat, because the most common workflow failure is pointing Claude at a vague objective and getting a vague result back.

A typical scenario involves a recurring process such as a weekly status report that runs four steps: collect updates from five teams, consolidate them, draft a leadership summary, and format the document. Claude fits naturally at consolidation and drafting, where the raw material exists and the work is synthesis, while collection depends on relationships and final review depends on accountability. The blueprint calls this solution design: placing Claude where it multiplies effort and keeping people at the steps where judgment or ownership matters.

Communicating value and limitations to stakeholders is the distinctive topic here. A certified associate can explain what Claude will reliably improve, where its output needs review, and why the same prompt can produce different results on different runs, all without overselling. The exam presents stakeholder conversations (a manager expecting perfection, a skeptical colleague) and asks for the framing that is both honest and useful.

What to study: map one of your own recurring processes end to end and decide where Claude belongs in it. That single exercise covers most of what this domain asks.

Domain 3: Governance, Risk, and Responsible Use (15%)

Approximate question count: ~9 of 60

Domain 3 tests the judgment your employer cares about most: knowing what belongs in an AI tool and what stays out. Appropriate use cases share a shape, with Claude drafting, summarizing, and brainstorming while a person reviews and decides, and the inappropriate ones share a shape as well: outputs shipping unreviewed, Claude serving as final authority on legal or medical questions, or decisions about people delegated to a model. The exam presents borderline cases and asks where the line falls.

Data sensitivity is the sharpest edge. Before pasting anything into a chat, a certified associate asks whose data it is, how sensitive it is, and what rules govern it. Customer personal information, employee records, health details, and confidential contracts carry obligations that exist no matter how convenient a quick summary would be, and privacy regulations apply to AI tools the same way they apply to email. The exam also rewards stripping identifying details when a task only needs the structure of a document.

Organizational AI policies get their own blueprint line, and the tested attitude is that policy governs use even when the tool technically permits something. The ethics topics close the domain: transparency about AI involvement where it matters, fairness whenever outputs affect people, and accountability, since the person who ships the work owns the work, whatever tool drafted it.

What to study: Anthropic's usage policies, the privacy material in the claude.ai Help Center, and your own company's AI policy, which doubles as realistic practice text for policy-application questions.

Domain 4: Prompting and Task Execution (14%)

Approximate question count: ~8 of 60

Domain 4 covers the craft most candidates assume the exam is mostly about, and its moderate weight reflects the blueprint's view that prompting serves the judgment domains above it. An effective business prompt supplies the context a task needs: audience, goal, output format, and constraints on length or tone. The exam repeatedly pairs a vague request (write a launch plan) with a specified one (product, audience, channels, timeline, format) and asks which missing element caused the generic result.

Task decomposition is the power tool of the domain. A quarterly business review goes badly as one giant prompt and goes well as a sequence: an outline first, then one section at a time with the relevant background pasted in, then a consolidation pass for consistency. Iteration follows the same logic, because when a draft misses, the productive move is a targeted follow-up naming what was wrong and what should change.

The blueprint closes with strategy adaptation across four task families. Brainstorming rewards open prompts and a request for volume before quality, analysis rewards explicit criteria and the data to apply them to, drafting rewards audience, tone, and an example of work you consider good, and research rewards a defined scope plus an instruction to separate established facts from interpretation.

What to study: the prompting guidance in the claude.ai Help Center, plus one rewrite exercise a day: improve a prompt you actually sent with context and format, then compare the two results.

Domain 5: Product and Model Selection (12%)

Approximate question count: ~7 of 60

Domain 5 tests whether you can pick the right part of claude.ai for the job. A plain chat fits a one-off question. Projects (workspaces with standing instructions and a library of uploaded knowledge files) fit recurring work that reuses the same background. Research mode (which has Claude search, read, and synthesize multiple sources into a cited report over several minutes) fits questions that need current, sourced information, and Artifacts fit deliverables you will edit and export.

Model selection applies the same proportionality across the three tiers: Haiku is the fastest and lightest, Sonnet the balanced middle, and Opus the most capable and most deliberate. Summarizing meeting notes is a bounded task where the source material is supplied and every tier lands close to the same quality, so Haiku handles it quickly, especially across a stack of transcripts. Stress-testing a market entry strategy rewards depth of reasoning, and Opus earns its slower pace there by catching the second-order problems a lighter model glides past. The exam treats reflexively choosing the most powerful model as a wrong answer when the task is simple.

Context limitations complete the domain. The context window is the model's working memory for a single conversation, and very long sessions degrade: earlier details fade, instructions blur, and answers drift. Know the three remedies: restart fresh when the accumulated history has become noise, ask Claude for a summary to carry into a new session when the history still matters, and move genuinely recurring material into a Project so it persists without being repeated.

What to study: the feature documentation in the claude.ai Help Center, plus one experiment: run the same summarization task on different model tiers and compare quality against the time each took.

Domain 6: Configuration and Knowledge Management (12%)

Approximate question count: ~7 of 60

Domain 6 turns individual skill into repeatable setup. Retyping the same background every session wastes time and produces inconsistent results, so the blueprint centers on Projects configured once and used daily: custom instructions that shape every response in the workspace, plus knowledge files that give Claude the source material it should treat as ground truth.

The scenario to internalize is a shared team Project. A marketing team would give its Project instructions covering brand voice, tone rules, terminology to use and avoid, and the default audience, then load its knowledge with the brand guidelines, current product one-pagers, and two or three approved campaigns as examples of the standard. Every draft any teammate requests then starts from the same foundation. Exam questions present a team need and ask what belongs in instructions (durable rules about how to respond) versus knowledge (source material to draw from), a distinction the blueprint treats as fundamental.

Connectors extend the idea to live data: a connector links Claude to a tool you already use, such as Google Drive or Gmail, with your permission, so it reads current documents instead of stale uploads. The blueprint pairs connectors with maintenance, because configurations decay as products rename, guidelines update, and teams change. A Project that still describes a discontinued product line is worse than no Project at all, since it injects confident, outdated context into every conversation.

What to study: build one real Project for a recurring area of your work, with instructions and at least three knowledge files, and note what you change after a week of use.

Domain 7: Troubleshooting and Optimization (10%)

Approximate question count: ~6 of 60

Domain 7 is the smallest domain, and its six questions test a repair mindset. When a prompt underperforms, the blueprint expects a diagnosis before a fix: is the instruction ambiguous, is necessary context missing, has the conversation grown too long and degraded, or is the task on the wrong feature or model tier entirely? Each cause has a different remedy, and the exam punishes reflexively rewriting the prompt when the real problem is a bloated session or a mismatched tool.

Adjusting from feedback is tested as a method: change one variable at a time (add the missing context, or tighten the format instruction, or start a fresh chat) and compare results, because changing everything at once leaves you unable to say what worked.

Optimization extends the same loop to workflows that already function: consolidating a six-prompt process into three, moving stable background into a Project, and saving the winning prompt as a template the whole team can reuse.

What to study: keep a short log of prompts that failed and what fixed each one. A week of real troubleshooting covers this domain better than any reading.

Cross-Domain Habits: How the Exam Thinks

CCAO-F questions rarely stay inside one domain, and the multiple-response items in particular love to span two or three. Three habits run underneath the whole blueprint, and recognizing them lets you answer questions you never explicitly studied.

Verify in Proportion to the Stakes

Verification depth (Domain 1), review requirements (Domain 3), model choice (Domain 5), and stakeholder promises (Domain 2) all scale with what happens if the output is wrong. When a scenario mentions a client, a contract, a regulated topic, or a decision about a person, the correct answer includes a human check.

The Person Owns the Work

Accountability never transfers to the tool: outputs ship under your name, policies bind you regardless of convenience, and stakeholders hear honest limitations from you. Whenever an answer option quietly removes the human from a consequential decision, treat it as a trap.

Set Up Once, Benefit Daily

The product-knowledge domains reward turning repeated effort into standing configuration: Projects over repeated pasting, the right feature over brute force, templates over reinvention. When two options both work, the exam prefers the one that makes tomorrow's work easier.

How to Prioritize Your Study Time

Weight is the starting point for prioritization, and experience is the adjustment. Here is a practical tiering for most candidates:

Start with Tier 1, because the verification mindset scores points far beyond its own 13 questions: it shapes correct answers in governance, stakeholder, and troubleshooting scenarios alike. Give Tier 2 the bulk of your scenario practice, since its 27 questions turn on judgment that develops by working through cases. Tier 3 moves fastest for regular claude.ai users, but run practice questions there early to confirm the assumption, since details like the instructions-versus-knowledge distinction trip up casual users. The 3-week CCAO-F study plan turns this tiering into a day-by-day schedule, and the first-attempt guide covers exam-day tactics, including the multiple-response items.

Common Questions

Domain Mastery Checklist

Use this checklist to confirm coverage before you sit the exam. Every line maps to a topic the blueprint names explicitly.

Conclusion

The CCAO-F blueprint tells you exactly what Anthropic believes a certified Claude associate does: judge and verify outputs (21%), design Claude into real workflows (16%), use it responsibly (15%), prompt it effectively (14%), choose the right product and model for each task (12%), configure it for repeatable quality (12%), and fix it when it underperforms (10%). Study in that order of weight, adjust for your own gaps, and give Domain 1 the respect its weight demands, because its habits echo through every other domain.

Three habits separate passing candidates. They match verification effort to the stakes of the output, they keep a person accountable for every consequential decision, and they reach for standing setup over repeated manual effort whenever a scenario allows it.

Your recommended next steps:

1. Build your schedule from the 3-week CCAO-F study plan, which allocates days according to the domain weights above
2. Read the first-attempt pass guide for question tactics, including multiple-response strategy
3. Keep the CCAO-F cheat sheet nearby for rapid review of the highest-yield facts
4. Weighing this certification against the developer track? The CCAO-F versus CCDV-F comparison settles which one fits your role

Ready to test yourself against the blueprint? Preporato's CCAO-F practice tests give you 6 full-length exams of 60 questions each, distributed across all 7 domains by weight, with an explanation for every answer. Take one cold, let the per-domain results set your study priorities, and book the real exam once your scores hold steady above the passing bar.