course contents
Claude Certified Associate – Foundations
CCAO-F mock exam
Our own items, written to the published exam guide. They are not real exam questions.
full mock
Readiness gauge only. The real exam reports a scaled score (pass at 720 on a 100–1,000 scale); percent-correct here does not convert to it. Aim for 75%+ in every domain.
practice by domain
sample questions
10 items from the mock bank, one per domain first. Open an item to see the answer and why.
An events officer at a university asks Claude to turn a list of 200 registrations into a file for the booking system. The system rejects any file whose column headers or date format differ from its import template, which uses the headers Name, Email and Session, with dates as DD/MM/YYYY. Claude's first file used its own headers and wrote dates as "3 March 2026".
What is the most effective change to the prompt?
Show the answer
Answer: C.
Correct: state the exact headers and date format, with one correct row. This is a technical task: the file either imports or fails, and the system accepts only its own template. The exact headers, the format and a sample row remove every guess. Why not the others: consistent dates — it would tidy the file, but one consistent style can still differ from DD/MM/YYYY, and the headers stay wrong; what booking systems usually need — Claude cannot see this system's template, so it guesses again; a self-check before replying — Claude can check only against what it knows, and it has never seen the template.
A school trust's finance director asked the operations manager, in a corridor conversation, for a summary of the trust's energy contracts. Claude has produced a fluent two-page draft from the contracts. Nothing was written down about what the summary must cover, and the director will use it to plan next year's budget.
What is the most effective first step in checking the draft?
Show the answer
Answer: A.
Correct: write the request down as a numbered checklist first. A missing topic leaves nothing on the page to notice, and the request exists only as a conversation, so there is nothing to check completeness against until it is written down. Why not the others: read closely for gaps — careful reading helps, but a draft read alone is checked only against itself, and an absent topic gives no sign; check each price and date — the right accuracy pass, but it cannot find what the draft leaves out, so it comes second; Claude's list of topics — it describes the draft, not what the director asked for.
A recruitment agency's consultant has one job advert to check against the agency's two-page inclusive-language guide before it goes live this afternoon. A colleague suggests setting up a Project for the guide. This is a one-off: the agency's adverts are normally checked by a specialist, and the consultant will not need the guide again this year.
What is the most proportionate way to do the check?
Show the answer
Answer: B.
Correct: a plain chat with both files attached. The check happens once, and the guide is short enough to attach, so a chat needs no setup and answers this afternoon. Why not the others: a Project with the guide — it would work, and it suits recurring work, but "the consultant will not need the guide again this year", so setting one up adds nothing; Research on inclusive adverts — it searches outside sources, while the rules that count are in the agency's own guide; an artifact edited in the panel — useful for a document to keep and share, but the job is a one-time check of wording.
A university's student services manager gave Claude notes from eight student focus groups and asked for the requirements of a new advice service. The list is tidy and quoted, and it says advice runs by appointment. Yet two groups asked for drop-in sessions and two asked for appointments. The director wants to see where students disagree before choosing either.
What is the most effective next prompt?
Show the answer
Answer: C.
Correct: list the disagreements, with a quote from each side. The notes split on drop-in against appointments, and the list settled it silently. The director wants "to see where students disagree before choosing", so the split belongs among the open questions, with both sides quoted. Why not the others: group number per requirement — it makes each line traceable, but the drop-in view is already missing from the list, so it still hides the split; Claude chooses what usually works best — it makes the director's decision from general practice; Research on other universities — useful later, but it says nothing about what these students want.
A law firm's office team on the Team plan is setting up a Project to draft engagement letters for new clients. Five files are on offer. The firm's policy keeps client personal data out of AI tools, and the office manager wants the knowledge to hold only what the letters need, since staff-only rules must never reach a client.
Which TWO files should go into the Project's knowledge?
Show the answer
Answer: C and D.
Correct: the current template and the current fee schedule. A Project's knowledge should hold only the current files its one job needs, and these two give every letter its structure, terms and rates. Why not the others: staff billing procedure — it is accurate and related, but it is written for staff, and the stem says staff-only rules must never reach a client; last year's template — an old version beside the current one gives Claude two sets of wording to mix; client contact spreadsheet — it puts client personal data into every chat, which the policy bars.
A hotel group's marketing coordinator is launching a new seaside hotel that has only six online reviews. The manager asks for fifty five-star reviews, written by Claude under made-up guest names, to post on booking sites before the summer. The coordinator still wants Claude's help to raise the hotel's real review count.
Which use of Claude best meets the goal while staying within Anthropic's Usage Policy?
Show the answer
Answer: A.
Correct: invite real guests to review. The Usage Policy lists fake reviews among the deceptive campaigns it prohibits, and no manager's request can approve them; Claude can still help with the honest route, an invitation that real guests answer. Why not the others: guest comments posted as reviews — the words are real, but the guests never posted them, so the reviews are still made up; a manager checks each one — a check can make a review accurate, not genuine; posted over several weeks — spacing them out hides the fake reviews rather than ending them.
A college's admissions team shares a saved prompt that turns interview notes into a one-paragraph summary of each applicant for the selection panel. The summaries are too generic to tell applicants apart. Two coordinators each patched the prompt, one adding "be specific" and one adding a role, and each patch seemed to help one week and not the next.
What is the most effective next step?
Show the answer
Answer: C.
Correct: name the cause, then change one thing on the same notes. Generic output can come from a missing audience, a missing example or an unclear task, and the patches were guesses, so each "seemed to help one week and not the next". Tracing the symptom to its cause, then re-testing on the notes that failed, shows what works. Why not the others: one example summary — a common fix that may well be right, but it is a third guess, judged on new notes; every known fix at once — nobody can tell which change helped; the most capable model — any model still guesses what the prompt leaves open.
A retailer's customer service lead types: "Summarise the complaint below and list what the customer wants." Under it they paste the customer's email, which ends: "Please reply by Friday with your decision on a refund." Claude's answer is a polite reply to the customer that promises a decision by Friday, with no summary and no list.
What is the most likely root cause?
Show the answer
Answer: A.
Correct: no boundary around the pasted email. The email ends with its own request, and nothing showed Claude where the material stopped and the instruction applied, so it acted on the email's request. Wrapping the email in tags such as
<email>separates material from instruction. Why not the others: email too long — one email fits easily, and Claude used its closing line; no length limit — a limit changes how long the text is, not whether Claude writes a summary or a reply; no format named — the prompt named two outputs, a summary and a list, and neither came back.A retailer's finance assistant asked Claude to summarise sales for the financial year April 2025 to March 2026, attaching a sheet with two total columns: calendar year 2025 and financial year 2025–26. The summary says sales were £3.1 million, which matches the calendar-year column exactly. The financial-year total in the sheet is £3.4 million.
What is the most likely root cause?
Show the answer
Answer: B.
Correct: the right figure from the wrong period. £3.1 million is exactly the calendar-year value, so Claude read the sheet but answered a different period from the one the request meant. An accuracy check must compare the period as well as the value. Why not the others: rounding — £3.4 million cannot round to £3.1 million; sheet too large — a sheet with two total columns is small, and Claude read the column beside the right one; an invented figure — an exact match with another column of the same sheet points to a misread, not an invention.
A lettings firm's compliance team checks 30 new tenancy agreements each month against its 12-point checklist, and three officers take turns to run the check. The result is a one-page findings report that the manager edits and exports as a PDF. Last month the officer on duty pasted an old version of the checklist, so the checklist must now reach every check without anyone pasting it.
Which TWO feature choices best fit this work?
Show the answer
Answer: A and B.
Correct: a shared Project with the checklist, and an artifact for the report. The checklist in project knowledge reaches every chat in the Project, so no officer has to paste it; the report is a document to edit, keep and export, which is an artifact's job. Why not the others: paste from a shared folder — it fixes the old-version problem, but it still depends on someone pasting, which the stem rules out; Research for each check — the test is the firm's own checklist, not outside sources; a long chat reply — it gives the findings, but copying them by hand loses the editing and PDF export.