PDF Summarizer

How do I summarize a PDF?

Press Choose a PDF, pick the summary type and length, and press Summarize PDF. The tool reads the text of every page, ranks the sentences or hands them to the browser's built-in AI, and shows the result with the engine named. Copy it, or save it as a Markdown or text file.

1. Press Choose a PDF.
2. Pick Key points, a Short paragraph, a Teaser or a Headline, and a length.
3. Choose The whole document, Each section or Each page.
4. Press Summarize PDF, then copy the result or save it as .md or .txt.

Is it private to summarize a PDF here?

Yes. The text is read in your browser with PDF.js, and both engines run on your device. The basic engine is plain code in this page. Chrome's built-in AI is a model stored on your computer. No PDF text goes to a server or to a cloud AI service.

Scanned pages are read with OCR in the same way, on your device. After the page has loaded, the basic summarizer needs no network. Chrome downloads its own model once, from Google, before the built-in AI can run; that download contains no data from your PDF.

How accurate is a PDF summary?

It depends on the engine. The basic engine picks the sentences that best represent the whole text, so every line is a true quote from your PDF, though it can miss a point or repeat one. Chrome's AI rewrites in its own words and can get a detail wrong, so check the numbers.

We summarized a real 6-page guide (1,302 words, 147 sentences, 8 headings) with the basic engine. Ranking all 147 sentences and choosing the key points took 24 ms on top of reading the text.

The medium key points chose these sentences, in the order they appear in the guide:

• A PDF is large because of what it stores: pictures with more pixels than the page needs, scanned pages saved as photos, fonts embedded in full, or old versions kept after edits.
• In our tests, the same A4 page of text was 3 KB as text, 288 KB as a 150 dpi scan and 1.9 MB as a 600 dpi scan.
• In our tests a page of text took 3 KB, while the same page scanned took 288 KB to 1.9 MB.
• Our A4 test page grew from 288 KB at 150 dpi to 1.9 MB at 600 dpi, while the same text as real text took 3 KB.
• Reduced to what an A4 page shows at 150 dpi, the same picture made 216 KB, and at screen size 101 KB.

With Each section, the same guide gets one sentence per heading:

Bullets two to four say nearly the same thing about scan sizes. That is the main weakness of an extractive summary: it ranks sentences one at a time, and a point the text repeats can come back more than once. The Short setting keeps fewer sentences, so it repeats less.

What is the difference between extractive and abstractive summaries?

An extractive summary picks the key sentences from the text and keeps them word for word. An abstractive summary writes new sentences, as a person would. The basic engine here is extractive. Chrome's built-in AI is abstractive, so it reads more smoothly but can state something the PDF does not say.

The basic engine splits the text into sentences, weights each word with TF-IDF (a rare word counts for more), compares sentences by cosine similarity and ranks them with TextRank, the method of Mihalcea and Tarau (2004). It keeps the top sentences in their original order and skips a sentence that nearly repeats one already chosen.

Which browsers have built-in AI summaries?

Desktop Chrome 138 or later has the Summarizer API, according to Chrome's developer documentation. Firefox and Safari do not, and phones are not supported yet. The computer needs a lot of free storage and memory, and the first use downloads the model. Where it is missing, the basic engine runs instead.

Chrome's page lists the requirements: at least 22 GB of free space on the drive that holds your Chrome profile, a graphics card with more than 4 GB of video memory or a processor with 16 GB of RAM and 4 cores, and a network connection for the first download. Its summarizer reads English, Spanish, Japanese, German and French. This tool checks all of that for you and says which engine it will use before you start. See Summarize with built-in AI in Chrome for Developers.

Can I summarize a long or scanned PDF?

Yes to both. A long PDF can be summarized section by section or page by page, so each part gets its own short summary. A scanned PDF has no text to read, so press the OCR button the tool shows; it reads those pages on your device first, then summarizes them with the rest.

Chrome's model reads a limited amount of text at a time, so a long document goes through in pieces and the pieces are summarized again. With headings found in the PDF, Each section uses them as titles; with fewer than two headings, the tool groups by page and says so. OCR reads the scanned text in the page's language when it is English, Spanish, French, German, Italian, Portuguese or Russian, and makes mistakes on poor scans. For a searchable PDF, use OCR PDF.