Ask questions about a PDF and check the cited pages
Open the document, go to the Ask tab of the AI assistant, type a question and click Ask. Answers come only from that file and list Sources: click a p. 4 chip to jump to the quoted page. Each message costs 1 credit.
Updated 2 min read Counts toward ai credits
On this page (3)
Step 1: Open the Ask tab
In the editor, click the yellow ✨ AI assistant button in the top bar (or More (⋯) → AI assistant). From the dashboard, click ✨ AI assistant → Chat with a PDF and pick a document. The panel opens on Ask this document.
Step 2: Ask a question
Type in Your question, or click a suggestion like What are the key dates and deadlines?. Click Ask.
Step 3: Read the answer
The answer streams in. Page numbers in the text and the chips under Sources link to the pages it quotes.
Step 4: Check the page
Click a p. 4 chip (Go to page 4). The editor scrolls to that page so you can read the quote in context.
- Send the question
- Enter
- New line in the question
- ShiftEnter
How citations work
Every quote in an answer is checked word for word against the text of the page it names. Quotes that don't match are removed, and the answer tells you how many ("1 quote couldn’t be matched word-for-word to the page and was removed"). If no quote survives, you'll see No verified page quote for this answer — check it against the document.
If the answer isn't in the document, it says Not found in this document. instead of guessing. When an answer was drafted but couldn't be verified against the page text, it is withheld and says so.
Stopping and failures
Click Stop while an answer is streaming to cancel it. A stopped or interrupted answer shows Nothing was charged. with a Try again button.
Scans, photos and images
A document with a text layer is always answered from its text. A short file with no text at all (a scan, a photo converted to PDF) is read from page images: up to 5 pages on Free and 10 on Pro and Business. Those sources show Read from the page image — no text layer to quote. Longer scans need OCR first; use the Scan tab or OCR PDF.
To ask about a photo, screenshot or chart, use Chat with an image. The image is saved to your documents as a one-page PDF and answers cite it as page 1.
Can I ask across several PDFs at once?
Yes, in a workspace. Add the documents, then use Ask across documents. See Ask across several documents.
The answer ignores the end of my document.
Your plan reads up to 100 pages per document on Free (600 on Pro, 2,000 on Business). The panel lists pages that weren't included.
Is my chat saved?
In Standard AI mode, yes. Turn off AI history in Data & privacy if you don't want that; see What happens to your documents when you use AI.
Tools in this article
Did this answer your question?
Related articles
- Summarize a PDF, search it by meaning, or map itIn the AI assistant, the Summary tab writes key points that link to their pages (2 credits), and Search finds passages by meaning (1 credit). For a visual overview, PDF to mind map builds a map whose nodes quote their pages.
- Ask across several documents and make briefs, slides or quizzesCreate a workspace in Workspaces, add documents, and use Ask across documents (2 credits per answer) or Make something from these documents to generate a brief, slide outline, quiz or audio brief. Every citation names the document, version and page.
- How accurate AI answers are, and how to check themAI answers can be wrong or miss nuance, so every result links to the pages it came from and chat quotes are checked word for word against the page. Check numbers, dates and obligations against the cited page before you rely on them.
- Make a scanned PDF editable and searchable with OCROpen the OCR PDF tool, upload the scan, pick the Languages and pages and click Recognize text. The result looks identical but has a searchable text layer; click Open in the editor to change its text. OCR is counted per page recognized.