Why specify a length?

Without a target, models default to whatever feels right and lengths drift between runs. A concrete length — a word count, sentence count, or bullet count — makes summaries consistent and lets you fit them into a fixed space like a digest or a card.

What is the difference between extractive and abstractive?

An extractive summary pulls key sentences verbatim from the source; an abstractive one rewrites the ideas in fresh wording. Abstractive reads better and compresses more, while extractive preserves exact phrasing — useful for legal or technical text where wording matters.

How do I stop it dropping the parts I care about?

List the key aspects to preserve. The builder turns them into an explicit instruction so the model keeps decisions, action items, figures, or caveats even when squeezing the length, instead of summarizing them away.

Does the content type really matter?

Yes. A meeting transcript needs decisions and owners; an article needs the thesis and evidence; code needs purpose, inputs, and side effects. Picking the type loads the right defaults so you are not configuring everything by hand.

Summarization Prompt Builder

Summarization prompt builder

“Summarize this” is one of the most common LLM requests and one of the most underspecified. The model guesses the length, the audience, and what is worth keeping — so you get a different shape every time and important details quietly vanish. A good summarization prompt fixes the length, names the reader, picks an abstraction level, and pins the points that must survive. This builder assembles that prompt for documents, meetings, articles, or code.

How it works

You choose a content type, which loads sensible defaults for what matters in that material. Then you set the target length (words, sentences, or bullets), the audience so the vocabulary and depth fit the reader, the abstraction level (extractive verbatim vs abstractive rewrite), and the key aspects to preserve — decisions, action items, figures, caveats. The tool writes these into a single prompt with clear, ordered instructions, plus an output-format directive so the summary lands as prose, bullets, or a structured digest you can drop straight into your workflow.

Tips and example

Name the reader. “For an executive who missed the meeting” produces a very different, more useful summary than an unspecified audience.
Preserve numbers explicitly. Models love to round or drop figures when compressing. If the metrics matter, list “key figures” as an aspect to keep.
Use extractive for legal or technical text. When exact wording carries meaning, ask for verbatim key sentences rather than a paraphrase.
Pick bullets for digests. A fixed bullet count makes summaries uniform and easy to scan across many items, which prose lengths never quite achieve.