Text Summariser
Pull out the sentences that carry the most of a document — extraction only, nothing is rewritten.
About this tool
Scores every sentence by how much of the document's own vocabulary it carries, and returns the highest-scoring ones in their original order.
**Every sentence in the output is copied from the input, word for word.** Nothing is rewritten or paraphrased and no language model is involved. That makes the result read more jumpily than a summary somebody wrote — and gives it one property a generated summary cannot offer: it is incapable of claiming something the source did not say.
Useful for getting the shape of a long page quickly, for checking whether an article actually says what its headline promises, and for finding the sentences a page would lose least by cutting.
Common questions
Does this use AI?
No. It ranks the sentences already in your text and returns the best-scoring ones unchanged. Nothing is sent anywhere and nothing is generated.
Why does the summary read a little disjointed?
Because it is made of sentences lifted from different parts of the document, and some of them refer back to things that were cut. That is the inherent cost of extraction; the benefit is that it cannot invent anything.
How are sentences scored?
By the frequency of the significant words they contain, divided by length so long sentences do not win automatically, with a small bonus for opening and closing sentences. Common words are ignored.
Can I use the output as a meta description?
Read it first. An extracted sentence is often close, but a meta description is a promise about the page and is worth writing deliberately. Use the top sentence as a starting point, not as the finished text.