Tools / Ask a PDF
Ask a question about a PDF
Answers from your document, with the page it came from.
- Your file is uploaded to do this
- Deleted 24 hours after processing
- What we do with it
Ask a PDF runs on a paid model, so it is part of Pro rather than the free tier. Every other tool here is unaffected. See the plans.
Choose a PDF
or drop them here
How it works
- Add the PDF.
- Type your question.
- Read the answer, then check the pages it cites.
Asking a long document one question
Reading forty pages to find one number is the thing this avoids, and the citation is what makes the answer usable.
Ask for a specific thing
What is the notice period, rather than tell me about this contract. A narrow question gets a checkable answer.
Always open the pages it cites
The citation is the point. An answer you have not verified against the page is not evidence of anything.
It reads this document, not the world
It answers from what is in the file. It is not a substitute for advice about what the document means for you.
Ask follow ups rather than one long question
Several narrow questions are easier to check than one answer covering five things.
Where this is the wrong tool. This uploads, and the document is read by a language model. The file is sent over an encrypted connection and deleted after 24 hours. Answers can be wrong or incomplete, which is why every one comes with the page it came from. For anything legal, financial or medical, check the source and ask a person.
When the browser is not enough
Scanly for Android
Take document-scanning workflows to your Android phone with Scanly. See the Google Play listing for current features, requirements and data-processing information.
Questions
Does it just make things up?
It is told to answer only from the document you gave it and to say 'That is not in this document' when the answer is not there. Every claim has to carry the page it came from, and the page shows you those page numbers separately from the answer. If an answer cites no page, treat it as an answer that has not shown its working.
Will it read a scan?
No. This tool requires an existing text layer. Image-only PDFs are uploaded before the server can reject them. For plain text from a scan, use PDF to JPG and then Image to Text. That workflow does not create a searchable PDF for this tool; use a separate searchable-PDF OCR workflow first.
What about a very long document?
There is a limit to how much it can hold at once. Past that it tells you which pages it read rather than quietly answering from the first part and letting you assume it read all of it.
Does my document leave my device?
Yes. The text is pulled out on our servers and the question and that text go to the model. Your upload is deleted 24 hours later. Do not use it for anything you are not willing to send.
Does it remember my last document?
No. Each question is answered from the document in that request and nothing else. There is no history and nothing is learned from your files.