Skip to content
Paragraph Generator

Drop a PDF here

Up to 20MB. The PDF needs selectable text, not scanned pages.

5 tokens to prepare a document

Chat with a PDF

  1. 01

    Upload

    Drop in a PDF up to 20MB. The document is read, split into passages and indexed, which takes a few seconds.

  2. 02

    Ask

    Type a question the way you would ask a colleague. The relevant passages are found first, then the answer is written from them.

  3. 03

    Check the pages

    Each answer carries the sections and page numbers it drew on, so you can jump to the source and confirm it.

Questions worth asking

Anything you would put to someone who had already read the file. Ordinary phrasing works better here than the keyword-style terms you would type into a search box.

Why bother

What you save by asking

Most of the time you do not want the whole document. You want one figure, one clause, or the shape of an argument, and reading forty pages to reach it is the expensive part.

Minutes rather than an afternoon

Ask what you came to find out and read the answer. The scrolling and scanning is the part this removes.

The page comes with the answer

Answers name the sections they were built from, so checking a claim is one jump instead of a fresh hunt through the file.

Ask again in plainer words

If an answer goes over your head, say so. You can have the same passage explained more simply without starting the whole thing over.

Nothing to set up

Drop the file in and type. There are no settings to work through, and no account needed before your first document.

Who it helps

For anyone facing a long document

Students

Find the argument in a paper without reading forty pages to locate one section.

Anyone reading a contract

Ask what a clause says and see the page it sits on before you rely on the answer.

Researchers

Interrogate a report for the figure or method you need instead of scrolling for it.

Busy readers

Get the shape of a document quickly, then go deep only where it matters.

Working with more than one document

Uploads stack up in the rail beside the chat. Each PDF keeps its own thread, so the questions you asked about a lease stay with the lease and never turn up in the conversation about a research paper. Switching between them is a click, and the earlier answers are still there when you come back to one.

The trade is that every answer is built only from the document it belongs to. A question that spans two files has to be asked twice, once in each thread, and the two answers compared by you. That is deliberate: it is the reason a citation can point at a real page rather than at a pile of documents in general.

When you are finished with a file, delete it. That removes the document and its conversation together, which is worth doing as a habit if you have been reading something you would rather not leave sitting on a server.

Scanned pages will not work
The document needs selectable text. A PDF made of page photographs has no text layer to read, so run it through OCR before uploading.
Turn on reasoning for hard questions
The reasoning toggle shows the model working through the passages before it answers. It is slower, and worth it when the answer needs to be defensible.
Other tools

Once you have found what you needed

Reading a document is usually the start of writing something. Every tool here runs on the same daily allowance, so moving a passage between them costs you nothing extra.

PDFs, up to 20MB, with selectable text. Scanned documents and photographs of pages have no text layer and will be rejected with an explanation rather than a silent failure.
Every answer lists the sections and page numbers it was built from. That is the point of the citations: you can open the document at that page and check the claim yourself.
Yes. Each upload opens its own thread, and they sit together in the rail beside the chat, so you can move between a contract and a report without either conversation bleeding into the other. Each answer is drawn from one document only, so a question about two files has to be asked in both threads.
Not directly. It answers questions about the document you uploaded, so what comes back is drawn from the file rather than composed from scratch. Take the points you found here to the paragraph generator or the expander when you want to write the piece itself.
No, and deliberately so. It reads the document in front of it, which is what makes it possible to name a page for every answer. Questions the PDF does not cover are outside what it can help with. For open-ended questions, use the AI Chat or the free ChatGPT page instead.
That is a separate job. This tool reads a document and answers questions about it; it does not edit it. Paste the passage into the grammar checker and it will flag the spelling, punctuation and word-choice problems and show you the corrected version.
Yes. Each document in the sidebar has a delete control that removes the file and its conversation permanently.
Your documents are tied to an identifier held in your own browser, not to your email address, and it is not guessable. Clearing your browser storage will lose access to previous uploads, so delete anything sensitive when you are finished with it.
Preparing a document uses five tokens and each question uses two, from the same daily allowance the rest of the site uses. The allowance refills daily and signing in raises it.

Put a document to work

Upload a PDF and ask it the question you actually have.