Calibrating Research Prompts and Source Discipline (CCAO-F)

Research is tight on scope and sources

Research sits near analysis on the control end of the dial, but its defining constraints are different: not evaluation criteria, but scope, currency, and source discipline. The Claude Certified Associate - Foundations (CCAO-F) exam treats calibrating a research prompt as an apply-level skill because a good research request is precise about what question is being asked, how wide the investigation should go, whether current sources are needed, and how claims should be grounded and cited.

Two judgements sit at the centre of this skill. First, matching the tool to the depth - a quick currency check versus a deep multi-source investigation are different jobs with different modes. Second, handling citations with discipline - asking for them so claims are checkable, while never assuming a confident-looking citation is automatically true. Both come down to keeping the research grounded rather than plausible.

Define the question, scope, and currency

A research prompt starts by fixing three things. The question: what precisely are you trying to find out, stated narrowly enough that the investigation has a target. The scope: the boundaries of what to cover - which competitors, which time frame, which markets - so the research stays bounded rather than sprawling. And the currency requirement: whether the answer depends on up-to-date information or whether established knowledge suffices.

That third judgement drives a tool choice. If the currency need is light - a quick check of a recent fact or two - turning on web search in chat is enough. If the task is a genuine multi-source investigation that needs breadth and synthesis across many documents, a dedicated deep research mode is the right instrument. Using a quick web-search-in-chat approach for something that actually needs deep investigation under-serves it, and using heavy research machinery for a one-fact currency check is overkill. Matching the depth of the tool to the depth of the question is part of calibrating the prompt.

Citations make claims checkable, but are not proof

Source discipline is where research prompts most often go wrong on the exam. Requesting citations is good practice: it makes each claim checkable and signals that the answer should be grounded. But a citation is a pointer, not a guarantee. A citation produced only from training memory - rather than from an actual retrieved source - can look every bit as confident and well-formatted as a real one while pointing at something that does not support the claim, or does not exist.

So the discipline has two parts. Ask for citations, preferably from a grounded source such as web search or research results, so claims can be traced. And treat the citations themselves with appropriate skepticism: a confident, tidy citation is a starting point for verification, not the end of it. For anything that matters, verify the important claims independently rather than trusting the format. The exam's point is that confidence and formatting are not evidence of grounding, and a well-presented citation can give false assurance if taken at face value.

What the CCAO-F exam trips candidates on

Two traps recur, and both concern grounding.

The first is treating a citation as automatically verified just because it looks confident and well-formatted. A scenario presents a polished citation and a distractor accepts it as proof. The credited reading is that a citation from training memory can look confident without being grounded - the disciplined move is to verify important claims independently rather than trusting the presentation.

The second is using a quick web-search-in-chat approach for a task that actually needs deep multi-source investigation, or the reverse. A scenario mismatches the tool to the depth, and a distractor keeps the wrong mode. The reliable reading is to match the instrument to the need: light currency checks suit web search in chat, while genuine multi-source investigation calls for a dedicated research mode.

Common misreadings to avoid

How this shows up on the exam

Domain 1 questions on this knowledge point show a research task with a mismatched tool or an over-trusted citation and ask for the disciplined response. The reliable frame is to define scope and currency, match web search or deep research to the depth needed, and treat citations as checkable pointers to verify rather than as automatic proof.

This knowledge point applies the four task types and the control-latitude dial to research, and it shares its verify-before-you-trust instinct with using verification tools during iteration for numeric accuracy, where the same principle applies to numbers. It is one of the four calibrations that feed into selecting strategy for a given scenario.