How to summarize a PDF with AI

The fastest way to summarize a PDF with AI is to open the file in a tool that reads PDFs directly — ChatGPT, Claude, Google NotebookLM, Microsoft Copilot, or Adobe Acrobat AI Assistant — attach or open the document, and ask for a summary. Each one takes the PDF a little differently, and each has limits on page count, file size, and whether it can read a scanned page. Pick the tool that matches your document, and read the privacy rule below before you upload anything.

The privacy rule that comes first#

Every method here sends your document to a company’s servers. Before you upload, ask whether the file is yours to share. Do not upload contracts, medical records, employee data, unreleased financials, or anything under an NDA or regulation to a general-purpose AI on a personal account. Consumer tiers often reserve the right to use your uploads to improve their models unless you opt out in the data settings, while business and enterprise tiers usually promise not to train on your content. When in doubt, redact the sensitive parts first, or use an on-device or approved enterprise tool. A summary is rarely worth a data-handling problem.

ChatGPT: attach the PDF and ask#

Open a chat, click the + (or paperclip) next to the message box, choose Add photos or files, pick your PDF, then type a specific request such as Summarize this in 10 bullet points, then list any deadlines. Per OpenAI’s documentation, each file can be up to 512 MB, but text-heavy files like PDFs also hit a cap around 2 million tokens — a very long report can be truncated or rejected even under the size limit. Free accounts get a few uploads a day; Plus and Pro raise that ceiling sharply. Because current ChatGPT models are multimodal, they can read a scanned PDF by looking at the page image, so a document with no text layer still works if the scan is clean.

Claude: strong on scanned pages and long reports#

In Claude, click the + button at the lower-left of the message box, choose Add files or photos (or drag the file in), then ask for the summary. Anthropic’s help center sets the clearest page rules of any tool here: for PDFs of 100 pages or fewer, Claude analyzes both text and visual elements like charts and scanned pages; from 101 to 1,000 pages it processes text only; and you cannot upload a PDF over 1,000 pages. Chat uploads allow generous size caps (up to 500 MB and 20 files per chat), but for most documents the page count is the real limit. The 100-page visual window makes Claude a good pick for scanned reports, tables, and handwritten notes, as long as the scan is legible.

Google NotebookLM: many documents, grounded answers#

NotebookLM is built for reading across a stack of sources. Open notebooklm.google.com, create or open a notebook, click Add sources, and upload the PDF from your device or Google Drive. It generates an overview automatically, and you can ask for a tighter summary in the chat box; every answer cites the source passage it came from. Per Google’s help, each source can hold up to 500,000 words or 200 MB, with 50 sources per notebook on the free plan and hundreds on paid tiers. The catch: NotebookLM reads a document’s text layer and does not run OCR on image-only PDFs, so a scanned or photographed file gives it nothing to index. Copy-protected PDFs will not import at all.

Microsoft Copilot: summarize in Edge or OneDrive#

Copilot gives you two easy routes. In Microsoft Edge, open the PDF, click the Copilot icon at the top-right, and pick the Summarize prompt; Edge 145 and later also add dedicated Summarize and Explain buttons directly in the PDF reader. For files in the cloud, Microsoft’s support page explains you can select a PDF at onedrive.com, click Copilot, and choose Summarize this file, or right-click the file in File Explorer and pick OneDrive then Summarize — up to 5 files at once. Copilot needs selectable text, so a scanned PDF will not summarize until it has been run through OCR.

Adobe Acrobat AI Assistant: summaries with page citations#

If your PDF already lives in Acrobat, open it in Acrobat or the free Acrobat Reader, click AI Assistant in the toolbar, and choose Generative summary. Acrobat streams the overview in chunks for long files and attaches clickable citations to the exact page and paragraph, so you can verify each claim. AI Assistant is a paid add-on billed on top of an Acrobat or Reader plan (it launched at $4.99 a month and is bundled into Acrobat Studio), so it is not free with base Reader. For a scanned document, run Acrobat’s own Scan & OCR > Recognize text first; the assistant then works, though quality falls on low-resolution or image-only scans.

Method comparison at a glance#

Tool	How you feed it the PDF	Size or page limit	Reads scanned PDFs?
ChatGPT	+ > Add photos or files	512 MB, ~2M tokens per file	Yes, via page-image vision
Claude	+ > Add files or photos	Up to 1,000 pages (visuals only ≤100)	Yes, up to 100 pages
NotebookLM	Add sources > upload	500,000 words / 200 MB per source	No — needs a text layer
Copilot (Edge/OneDrive)	Copilot sidebar or Summarize	Up to 5 files at once in OneDrive	No — needs selectable text
Acrobat AI Assistant	AI Assistant > Generative summary	Streams long PDFs in chunks	Only after OCR

When the file is scanned, locked, or too large#

Most failures trace back to three causes, and each has a fix. If a summary comes back empty or wrong on a scanned document, the PDF is a picture of text with no text layer — either use a vision-based tool (ChatGPT or Claude), or add a text layer first. Running the file through OCR, or converting the PDF to Word and re-saving it, gives the text-only tools something to read. If the upload is rejected for size, compress the PDF or split it into sections before trying again. And if a report spans several files, merge the PDFs into one so the tool summarizes the whole thing in a single pass rather than in disconnected pieces.

Next steps#

For a quick read, attach the PDF to ChatGPT or Claude and ask a specific question. For a research pile you will revisit, load it into NotebookLM so answers stay cited. For a file already open in your PDF editor or browser, Acrobat AI Assistant and Copilot are the shortest path. Whichever you choose, always name the output you want — a five-bullet summary, a list of action items, a plain-language explanation — since a precise prompt beats a generic one every time, and confirm any figure or date against the source page before you rely on it.