Tell Claude the Format You Want Back

A core prompting habit of telling Claude the shape you want the answer in — a table, a bulleted list, JSON, a two-sentence summary, a specific set of headings — rather than leaving the format to chance and reworking it after.

Without a stated format, Claude picks a reasonable default that may not match how you'll use the answer, so you spend a second turn reshaping it. Naming the format up front makes the first response fit its destination — a spreadsheet, a document, a code file. Claude follows format instructions reliably, so this is one of the cheapest ways to cut iterations.

Any time the answer has a downstream home: pasting into a doc, feeding a script, comparing options in a table, or when you want consistency across several answers.

Open-ended exploration or brainstorming, where a rigid format would cut off useful tangents. Let early thinking breathe; impose format when you're ready to use the output.

<Your request>.

Format the answer as: <e.g. a markdown table with columns X, Y, Z / a numbered list of at most 5 items / valid JSON matching this shape: {...} / three short paragraphs under these headings>. Follow the format exactly and don't add extra commentary outside it.

You ask for a competitor comparison 'as a markdown table with columns for price, key feature, and best-for'. Claude returns exactly that table, which you paste straight into your doc — instead of a wall of prose you'd have had to tabulate yourself.

For output a program will consume, give Claude an explicit schema or a filled example of the exact structure, and tell it to return only that with no prose around it. A concrete example constrains the shape far more reliably than a description of it.

• Asking for JSON but leaving the field names to Claude, so the shape varies between runs — pin the schema.
• Over-specifying format for creative work, flattening it into a checklist.
• Requesting a format and then also 'explain your reasoning', which reintroduces the prose you were trying to avoid — put reasoning in a separate turn.