Read
An AI PDF reader that reads the whole paper and cites the page
Most tools treat your document as an attachment to a chat window. Here the PDF is the thing you are in: the AI reads the whole paper before it answers, every quote opens on the page it came from, and your marks and its margin notes stay in your folder for next time.
- Opens
- PDFs in your folder — no upload, no copy
- Citations
- Page and sentence, checked first
- Runs on
- macOS (Apple silicon) · Windows
- Price
- Free app · AI plans from $10/mo, or your own key
- AI
- Included, or your own API key
You upload a file, ask a question, get a paragraph back, and the file goes away. Next week you start over. A PDF deserves better than that: it has pages, a claim about it is only as good as your ability to find the page it came from, and the work you do on it should still be there next month.
It reads the whole paper before it answers
What is this paper’s real contribution? Does the conclusion follow from its own results? What does it never test? No single passage holds those answers, so a keyword or vector search — which hands the AI the few fragments that best match your wording — cannot find them. Reading the paper can.
- A paper, report or chapter goes into the AI’s context whole — the budget covers most of them. Nothing is chunked, ranked or left out, so a question about the whole paper is answered from the whole paper.
- A book too long for that is read the way you would read it: the table of contents first, then the chapters the question needs, read back as whole sections. If that does not settle it, the AI goes back to the contents and opens another.
- PDFs that ship without bookmarks still get that table of contents — built from the file’s tagged headings, or failing that from heading sizes and numbering.
- That is also how it answers “what is missing?”: absence only shows up to something that read everything. Each answer lists what it actually read, so “it is not in the paper” and “it did not look” are never the same answer.
A quote lands on the page, not just in the file
“It says so in the report” is not a citation. Page 47, this sentence, highlighted — that is.
- Every quote from a PDF carries its page. Open the citation and the quote shows in its surrounding text; follow it and the reader goes to that page with the sentence highlighted.
- Quotes are verbatim and checked against the document. A quote that does not match the text is never shown to you with a straight face: an AI margin note is sent back, and an answer flags it at the end.
- Because paragraphs are rebuilt across page breaks, a quote is not cut in half by the bottom of a page or polluted by the running header above it.
The figure, the table, the equation: box it and ask, or take it with you
A paper outside your field rarely breaks on the prose. It breaks on Figure 3, on the column in Table 2 you cannot place, on the term in equation (7) nobody defines. Those are the parts a chat window built around text cannot see.
- Drag a box around a figure, a table or an equation and ask about just that. Its page and section go with the question, and so does the text under the box — so the numbers and labels come from the document, not from squinting at pixels.
- Or turn it into a note: a table becomes a real Markdown table, a chart becomes its data, an equation becomes LaTeX. The first menu item works out which it is.
- The note is written beside the PDF — paper.pdf gets paper - table.md — so the result tables from five papers can sit side by side in one comparison.
- When the explanation is what you wanted, save the region and the answer into your daily note in one step. The picture is copied into your own folder, so the note does not break when the conversation goes.
An AI that reads it with you, in the margin
Asking questions assumes you know what to ask. On a paper outside your field, you often do not.
- Pick a lens — guide, review, critique or study — and it works through the PDF chapter by chapter, leaving notes on the printed page beside the passages they are about.
- Ticks in the navigator show where the notes sit, so you can see the shape of a paper before you read a line of it.
- Stop it whenever you like and keep what it has done. The closing write-up says which chapters it is based on.
- Adopt the notes worth keeping as your own comments — they outlive any later pass, and they stay in your folder.
A scanned PDF becomes text you can ask about
The old textbook, the library scan, the photocopied proceedings: a PDF with no text layer is a stack of photographs, and an AI that answers about it anyway is answering from nothing.
- A scan is recognised as a scan the moment it opens. It reads as images, and the AI actions say why they are waiting instead of answering.
- Recognition is a button you press on this document. Long books run in batches, and pages already done are kept if the run is interrupted.
- Every page gets its own quality record, so the pages that came out badly are listed — and can be re-run on their own — instead of hiding inside a quotation you trust.
- The recognised text opens as a readable column with a “page N” marker wherever a page begins. Click one and you are on that page image; quotes the AI gives you land there too.
- Search it
- Highlight it
- Ask about it
- Quotes land on the page
The rest of the reader
Marks that stay where you made them
Highlights and comments are anchored to the real position in the file, not to an offset that drifts when the text is re-extracted, and they are kept in your folder, not on a server.
Select a passage, ask or explain
Ask about just that passage, or press Explain for a plain reading of it. Its page and section go with the question, and the answer cites back into the same page.
Ask across the papers in the folder
Put a whole folder of papers into one question — which ones use this method, where do their results disagree — and every quote still opens on its own page of its own PDF.
Read a two-column paper as one column
Switch to the text view: one readable column with real headings, paragraphs joined across page breaks, running headers dropped. A page marker takes you back to the original, and switching views keeps your place.
Contents, find, and where the notes are
A table of contents panel, find in the document, and ticks in the navigator marking every highlight and note.
Translate it, read it side by side
Translate the whole PDF with its terminology held steady from start to finish, then read the translation beside the original.
How translation works →Not only PDFsWord, PowerPoint, Excel and plain text files open in place too, with the same search and cited answers — supported formats lists what is read in each and where a citation lands. EPUB ebooks have a reader of their own.
Coming from a chat-with-PDF website: what changes, what stays
Most people who land here have dropped a PDF into a browser tab and asked it something. Here is what doing that here changes — including the rows where the browser tab wins.
| Topic | A chat-with-PDF websiteUpload a file, ask in a chat window | In MindMax |
|---|---|---|
| Where the PDF is | Uploaded to the service, kept on its terms. | The file in your folder, opened in place. Reading, marking and searching happen on your machine; when you ask, the text the AI needs goes to the model. |
| What an answer draws on | Varies by tool, and it is not always visible which parts of the file were used. | The whole paper when it fits; a long document by its table of contents. Each answer lists what it read. |
| Checking an answer | Often a reference you can click to the spot. | A verbatim quote, checked against the text first, opening on the page with the sentence highlighted. |
| Figures, tables, equations | Depends on the tool — worth checking whether it can see a figure at all, or only the text around it. | Box one and ask, with the text under the box sent along; or turn it into a Markdown table, chart data or LaTeX, saved beside the PDF. |
| Marking it up | Built around asking rather than marking. | Highlights, comments and AI margin notes on the printed page, kept in your folder rather than on a server. |
| Next month | A new chat, often a new upload. | The same file with the same marks and notes — and the other papers in the folder can be asked about together. |
| Getting started | Nothing to install. Open a tab. | A desktop app to download, then point it at a folder. |
| Phone and borrowed computers | Works in any browser. | Desktop only — macOS and Windows. |
| One question about one file | Hard to beat. | Works, but installing an app for a single file you will never open again is overkill. |
One part of a loop: read, think, write
Your notes, and the two hundred PDFs you read to write them, sit in one folder — where a single AI works across both. Tools you would otherwise stitch together each hold one half of that.
- ReadThis pageRead the paper here — the whole of it in the AI’s context, with quotes that open on the page.
- ThinkOpen a mind map beside the paper and lay its argument out as you go, or send a passage to a canvas card.
- WriteWrite notes or a draft in Markdown in the same folder, and quotes from the paper keep their page.
Where it gets hard
- A reader, not a PDF editor
It reads, marks up and answers questions. It does not edit page content, fill in or sign forms, or merge and split files — keep your PDF editor for those.
- On an ordinary PDF, the text view is for reading
Highlights, comments and AI annotations belong to the printed page, so you make them in the layout view. The text view is where you read and copy. On a recognised scan it is the other way round, because the page image has no text to mark.
- Complex layouts and heavy notation
Multi-column papers with interleaved figures, and dense mathematical notation, are the hardest cases for any extractor. Quotes from those pages are worth following to the page, and the notice in the text view — which says what the text leaves out and flags a thin extraction — is the first place a bad one shows.
- A whole book is not held at once
Past the budget, a long document is navigated by its table of contents and read back section by section. That is careful, but it is not the same as having every page in view — for a question that spans the whole book, name the chapters that matter.
- Scanned documents in poor condition
Recognition quality depends on the scan. MindMax reports per-page quality rather than pretending the whole book came out clean.
Common questions
Does it read the whole PDF, or search pieces of it?
The whole PDF, when it fits — and most papers, reports and chapters do. The full text goes into the AI’s context, so nothing is retrieved and nothing is missed. A long book is navigated by its table of contents instead, reading back whole sections, and each answer lists what it read.
When the AI quotes a PDF, can I check the quote?
Yes. Every quote carries its page. Open it and the PDF goes to that page with the sentence highlighted, and quotes are checked against the text before you see them.
My PDF has no bookmarks. Can I still navigate it by chapter?
Yes. A table of contents is built from the file’s tagged headings, or failing that from its heading sizes and numbering. The AI uses the same table of contents to find its way around long documents.
Can I read a PDF as plain reflowed text?
Yes. Switch to the text view and the extracted text is laid out as a readable column with its headings, with a small page marker wherever a new page begins. Click a marker and you are back on that page of the original layout.
Can I ask about a figure or an equation in a paper?
Yes. Drag a box around it and ask. The page, the section and the text under the box go with the question, so labels and numbers come from the document itself. You can also turn the region into a note — a table into a Markdown table, a chart into its data, an equation into LaTeX — written to a file beside the PDF.
Does reading PDFs cost credits?
No. Opening, reading, highlighting and searching your own PDFs never spend credits. Credits go to AI work — asking, explaining, annotating, translating — or you can use your own API key.
Do my annotations stay if I stop using MindMax?
Yes. They live in your own folder alongside the document and are anchored to the real position in the file.
Can I ask questions about a scanned PDF?
Yes, once it is recognised. A PDF with no text layer opens as images and the AI actions wait, saying why. Start recognition on that document and every page gets a quality record; after that you can search, highlight and ask about it, with quotes that open on the page.
See also
Whole-text answers, a folder as the subject of one question, and what each answer read.
Guide, critique and study — plus a review written to the author. What each one marks, and what it writes up at the end.
The same box in books, slides, images and notes — and what stops a figure being answered from pixels alone.
Recognition you start on purpose and can audit page by page.
The same surface at book scale, with chapter-level navigation.
The 30-minute workflow this page is built for.