Why Claude Writes So Much and How to Get Shorter Answers

Claude writes long responses because it was trained on feedback that rewarded thorough, well-developed answers. Without a length instruction, Claude defaults to giving more rather than less. It does not know when you would prefer a short answer over a long one unless you say so. Saving a set of response format instructions in fenbs means any connected assistant reads them at the start of a session and follows them without you repeating the request each time.

The fix is to tell Claude the shape of the answer you want before you ask the question. A length limit, a format, and a stopping rule together are usually enough.

Why Claude generates long responses

Claude was trained by having humans rate responses and giving higher scores to answers that were complete, thorough, and well-reasoned. Longer answers often scored better than shorter ones. Over many training cycles, this produced a model that defaults to thoroughness when no other instruction is present.

Claude also does not have a natural stopping point the way a human writer does. A human knows when a piece feels done. Claude keeps adding until it reaches the end of its current chain of reasoning or the message limit. Unless you define "done," Claude defines it for itself.

How to set a length limit

The most direct way to get a shorter answer is to state the length you want before asking the question. Word counts, sentence counts, and paragraph counts all work. "Answer in three sentences or fewer." "Write no more than 150 words." "Give me a one-paragraph answer."

Be specific. "Be brief" is vague and Claude will interpret it differently each time. "Answer in two sentences" is specific and Claude will follow it consistently. If you need a short answer on a complex topic, say so: "Give me the key point in one sentence, then stop."

• "Answer in two sentences or fewer."
• "Write no more than 100 words. Stop when you reach the limit."
• "Give me a five-bullet summary, one line per bullet."
• "One paragraph only. No preamble."

How to request a specific format

Format instructions reduce length by removing the prose that fills the gaps between the actual content. A bullet list is shorter than the same information in paragraphs. A table is shorter than a narrative comparison. An numbered list of steps is shorter than an essay about how to do something.

Use format instructions to match the output to how you will actually use it. "Give me this as a numbered list." "Summarize in a three-column table: issue, cause, fix." "Write this as five bullet points, no full sentences." Each format instruction shapes both the length and the structure.

How to stop Claude from adding summaries and preamble

Claude often adds a preamble before the answer ("Great question! Here is what you should know...") and a summary after it ("In conclusion, the key points were..."). Neither adds information. Both add length.

Tell Claude to skip them: "Answer directly. No preamble, no summary, no conclusion." You can also add "Do not say 'certainly' or 'great question' or 'in conclusion.'" Claude follows specific exclusions reliably.

How to get Claude to stop mid-answer

Sometimes Claude starts an answer and you want it to stop before it finishes. You can interrupt with a new message: "Stop here. Give me just what you have written so far." Claude will treat that as a new instruction and stop.

For tasks that tend to run long, a staged approach works better than waiting for Claude to finish and then editing: "Give me the first section only. Wait for my feedback before continuing." This breaks the output into reviewable chunks and prevents you from receiving a long answer you then need to cut down.

When longer answers are useful

Not every long response is a problem. When you are exploring an unfamiliar topic, working through a complex technical problem, or asking Claude to review a document in detail, a thorough response is the right response. The issue is length applied to tasks where a quick, direct answer is what you actually need.

Build the distinction into your instructions rather than applying the same format rule to every task. Keep a short-answer template for everyday questions and a detailed-review template for situations where completeness matters. When you switch tasks, load the right template.

How to do this with fenbs

1. Create a free account and set up a project for work where you regularly interact with Claude.
2. Open Skills and prompts and add a "Response format" entry with the length limits, format preferences, and exclusions that match your usual workflow.
3. Connect your AI apps from Integrations and ask the assistant to read your skills at the start of each session.
4. If you have different format needs for different tasks, save separate skills: one for quick answers and one for detailed reviews. Ask Claude to follow the right one for each task.
5. If a format rule is a team standard, add it to Decisions and rules so every connected assistant follows it.