A library FAQ must steer Claude toward a preferred citation style. Which prompting technique fits?

Steering a library FAQ toward a preferred citation style is efficiently done with a single-shot example of that style in the prompt. A librarian persona sets a helpful tone, but it does not communicate order, punctuation, italics, and required fields as reliably as a worked example, and tone alone won't stop models from conflating MLA, APA, and Chicago variants. Defining a librarian persona to establish a helpful tone fails because a persona does not specify punctuation, italics, or required fields, so house citation style will still drift across answers. Including a multi-shot set of contrasting citation formats fails conceptually—showing multiple styles dilutes the house style and does not teach which bibliographic format to use for FAQ sources. Instructing Claude to reason step by step through the citation elements fails because chain-of-thought helps with logic but does not enforce a strict house style, and it can still produce inconsistent punctuation and formatting rather than the strict house style patrons and staff expect. Exam caveat: the exemplar must match the exact style guide the library publishes, including digital-object identifiers, URLs, or municipal document quirks unique to local collections. Operational check: include one gold citation in the system prompt, run a battery of source types (book, article, city report), and reject FAQ answers whose citations deviate from a lightweight format checker.