User Guide
Add the PDF
The text is read straight out of the file. A scanned PDF has no text stored in it, so summarising a scan will return nothing — run it through the OCR tool first.
Choose the length
Brief returns three to five key points. Standard gives a comprehensive summary. Detailed keeps considerably more of the argument, which suits reports you intend to act on.
Choose the focus
General overview weights ordinary prose. Technical / scientific weights terminology and numbers, so methods and findings survive rather than being smoothed away.
Summarise
Everything runs on your own machine. There is no queue, no API call and no waiting on someone else’s server, so a long report takes seconds.
Read it as a map, not a replacement
The summary points you at the parts of the document that matter. For anything you will rely on — a contract, a medical report, a filing — go back and read those sections in full.
Keep it to one document
Summarising works on a single PDF at a time. To condense several, summarise each and then compare, which also keeps the sources distinguishable.
About the AI PDF Summarizer
About the PDF Summarizer
This tool reads the text of a PDF and returns the sentences that carry the most of its meaning. It runs entirely in your browser and makes no network requests at all once the page has loaded.
How the summary is produced — and what that means
It is worth being precise, because “AI summariser” covers two very different things.
This one is extractive. It scores every sentence in the document — on the significant terms it contains, how often those terms appear elsewhere, where the sentence sits in its section — and returns the highest-scoring ones in their original order. Every word in the summary is a word from your document.
The alternative is abstractive summarising, where a language model writes new prose describing the source. That reads more fluently, and it can also state things the document does not say. Extractive summarising cannot invent, because it never writes anything.
| This tool (extractive) | LLM (abstractive) | |
|---|---|---|
| Where the words come from | Your document, verbatim | Generated |
| Can it state something untrue? | No | Yes — the known failure mode |
| Reads like prose? | Less smoothly | Yes |
| Where your document goes | Nowhere | To a third-party server |
| Cost and speed | Free, seconds | Metered, slower |
For a confidential document the difference is not academic. Pasting a contract or a patient report into a hosted model sends it to someone else’s infrastructure. This does not.
The two settings
Length controls how many sentences survive. Brief is for triage — is this document about what I think it is? Standard is the working default. Detailed keeps enough to brief someone else from.
Focus changes which words count as significant. General overview treats ordinary vocabulary as the signal, which suits reports, articles and correspondence. Technical weights domain terms, units and figures, so a methods section is not discarded in favour of the introduction — useful for papers, specifications and clinical documents.
What it works on, and what it does not
It needs stored text. A born-digital PDF — exported from Word, generated by a system, printed to PDF — has it. A scan does not: each page is a picture, and there is nothing to read. The quick test is whether you can select a line of text with your cursor. If you cannot, run the file through OCR first.
It also assumes prose. A document that is mostly tables, figures or forms has few sentences to rank, and the summary will be correspondingly thin — not because it failed, but because the meaning of a table does not live in its sentences.
Using it well
Summaries are a filter, not a substitute. Use one to decide whether a document deserves your attention, to remind yourself what a file you read last month contained, or to find the section you actually need in something long. Read the original before you act on anything that matters.
Frequently Asked Questions
Does this send my PDF to an AI service?
No. It makes no network requests at all — the summary is produced on your own machine by scoring the sentences already in the document. Nothing is uploaded, which is the whole reason it is safe for confidential material.
Is the summary written by a language model?
No, and that is deliberate. It is extractive: every sentence in the summary is taken verbatim from your document. A language model writes new prose, which reads better but can state things the source never said.
Why did my scanned PDF produce nothing?
A scan is a picture of a page with no text stored in it, so there is nothing to summarise. Run it through the OCR tool first to recognise the words, then summarise the result.
What is the difference between General and Technical focus?
Which words are treated as significant. General weights ordinary vocabulary, suiting reports and articles. Technical weights domain terms, units and figures, so methods and results survive instead of being smoothed away.
How long should I make the summary?
Brief for triage — deciding whether a document is what you think it is. Standard for everyday use. Detailed when you need enough to brief someone else without reopening the original.
Can I summarise several PDFs at once?
No, one at a time. Summarising them separately is usually better anyway, since a combined summary loses track of which document said what.
Will it work on a document that is mostly tables?
Poorly, and unavoidably so — it ranks sentences, and a table has few. The information in a table lives in its structure rather than its prose.
Is the summary accurate?
Every sentence is genuinely from your document, so it cannot misquote. What it can do is omit context, since any summary is a selection. Read the original before acting on anything consequential.