How to chat with a PDF using AI
Updated
Contents
To chat with a PDF, open Chat with PDF, add your PDF, type a question in the Your question box, and press Ask the PDF. An AI model answers from the document's text and marks the pages it drew on as [p.N], and the box then becomes Ask a follow-up for your next question.
Use the answers to find your way around a document, not in place of reading it. The model is told to answer only from the document's text, to use no outside knowledge, and to say plainly when the answer is not there, yet nothing checks an answer against the pages afterwards, so an answer can still be wrong. Each answer carries a reminder to check it against your PDF.
Chat with PDF is one of the AI tools, so the work does not happen in your browser. With each question, your PDF uploads over an encrypted connection to our secure server, which extracts the selectable text and sends that text, with your question and the recent exchanges, to a third-party AI provider; the PDF file itself is never sent to the provider. The upload is deleted when the job ends, and our server's link to each answer stops working after 60 minutes. No account is needed.
Step by step
- 1Open the Chat with PDF tool. Ask the PDF works once the status badge reads Secure server.
- 2Drop one PDF on the box that says “Add your PDF, or browse”, or click it to choose the file. It needs selectable text; for a scan, run OCR PDF first.
- 3Type your question in the Your question box, up to 2,000 characters, in plain language: the notice period in a lease, say, or the total due on an invoice.
- 4Press Ask the PDF. Your PDF uploads over an encrypted connection to our secure server, and the button follows the job while the model answers.
- 5Read the answer in the conversation above the box. Where it shows a [p.N] marker, open that page of your PDF and check the wording before you rely on it.
- 6For another question, type in Ask a follow-up and press Send follow-up. Each answer has a button named after its ZIP that saves it to your device: lease.pdf gives lease-answer.zip.
What a [p.N] citation is, and what it is not
Before the model sees the text, our server marks each page that has text with its number, and the model is told to cite the pages it uses as [p.N]. The number is the page's position in the file, counted from the first page with blank pages included, so it can differ from the number printed on the page when a document opens with a cover or roman-numbered front matter.
A citation is the model's own pointer, not a quotation checked by software. The answer.md file in each download lists the cited pages, keeping only page numbers that exist in your file, but nothing checks that a cited page says what the answer says it does. Treat [p.N] as where to look, and read that page. An answer with no citation gives you nothing to look up, so search the document for it before you rely on it.
Follow-up questions and long documents
The conversation lives in your browser tab, and every question is a job of its own: the page uploads the PDF again and sends the most recent exchanges with it, up to 6. A note above the button says how many travel with the next question, and whether any were shortened or left out to stay within the request limits.
Our server does not link one question to the next. Reloading the page, choosing another file, or pressing Remove beside the file name starts a new conversation.
Each question reads a bounded amount of text, with your question and the conversation sent with it counted first, and the exact figure is stated on the Chat with PDF page. Past that point the model is told not to claim the document lacks something, and a notice above the answer says how far into the document the reading got. For a long file, pull the part you need into its own PDF with Extract Pages and ask that; its [p.N] citations then count the pages of the extracted copy. Files are capped at 100 MB.
The free allowance and where your text goes
Questions are free, within a daily allowance. 30,774 free characters of AI text a day for your connection. One run never counts more than 10,258 of them, however long the document is, and the allowance resets daily. Each question counts against the allowance as a run of its own, because this PDF's text and the conversation so far are sent again with every question. However long the document, at least 3 chat questions fit in a day if nothing else draws on it. The other AI tools draw on the same allowance. The charge is taken when the job is submitted, so a question our server refuses or that fails still uses it.
Our server sends the PDF's extracted text, with your question and the recent exchanges, to a third-party AI provider; the PDF file itself is never sent to the provider. The settings in our provider account that would let it use this text to improve its models are switched off, though the provider may keep what it receives for a limited time to watch for abuse. Because the same text goes again with every follow-up, the provider also holds it in a short-lived cache, measured in minutes rather than days. Section 3 of our privacy policy covers the provider's side.
Scanned PDFs need OCR first
The tool reads only text stored in the PDF as text. Your browser checks the file for text before anything is sent, and a scan with none is stopped there with a message pointing you to OCR PDF; that refusal takes nothing from the allowance.
Run the scan through OCR PDF with the document's language selected, then ask your questions of the searchable copy. Answers can only be as good as the recognised text.
Common questions
Can I trust the page citations?
Use them as pointers, not proof. The model is told to cite the pages it used as [p.N], and the download lists only page numbers that exist in your file, but nothing checks that the page says what the answer claims. Open the cited page and read it before you rely on the answer.
How many questions can I ask for free?
30,774 free characters of AI text a day for your connection. One run never counts more than 10,258 of them, however long the document is, and the allowance resets daily. Each question counts against the allowance as a run of its own, because this PDF's text and the conversation so far are sent again with every question. However long the document, at least 3 chat questions fit in a day if nothing else draws on it. The other AI tools draw on the same allowance.
Who sees the text of my PDF?
A third-party AI provider receives the extracted text, your question and the recent exchanges; the PDF file never reaches it. Our server deletes each upload when its job ends, and the link to each answer stops working after 60 minutes.
Does it remember my conversation?
Only in your open browser tab. The page re-sends the recent exchanges with each follow-up, and our server keeps nothing that links one question to the next, so reloading the page or removing the file starts over.
Should I chat with the PDF or summarize it?
Chat when you have a specific question, such as a date, a figure or a clause. For an overview of the whole document as paragraphs, bullet points or an outline, Summarize PDF draws on the same daily allowance.