Check the upload in your account rather than assuming every EPUB will work. OpenAI's supported-file page names common document formats, but does not explicitly name EPUB in its examples. That omission does not prove EPUB is always rejected. If your EPUB is accepted, verify its contents; if it is rejected, use a supported-format fallback.
This guide is about getting text into a conversation. For deciding which app to read in over several sessions, use the separate ChatGPT versus AI ebook reader comparison.
First, check the file and the result
- Open your EPUB in a normal ebook reader to check that the file is readable. Use a file you are permitted to upload.
- Try attaching it in ChatGPT. Available upload options depend on your account and usage limits.
- If processing succeeds, ask for the chapter headings and a short passage from the part you need. Compare them with the original.
- Ask a question about that section. If the answer is incomplete, work with a smaller excerpt instead of repeatedly asking for the whole book.
For example: “Using only the attached text, explain the claim in this section. Quote a short supporting passage and say when the evidence is missing.” This is a verification habit, not a guarantee that the model has processed every page.
If ChatGPT rejects the EPUB
For an unprotected file you can convert, a practical fallback is EPUB → TXT when you need the words, or EPUB → PDF when preserving page layout matters. The calibre conversion manual explains conversion settings. Inspect the converted file before uploading: check chapter order, footnotes and whether the relevant text survived.
Simply changing a filename from .epub to .pdf does not convert its contents. Conversion also does not remove every access restriction. If your book is tied to a store's reader, use a permitted text excerpt or another accessible copy rather than assuming any converter can open it.
Troubleshooting checklist
| Problem | A useful next check |
|---|---|
| Upload is refused | Review supported formats and current account limits; try a verified TXT or PDF conversion |
| “Upload limit reached” | Check account usage and storage; repeated failed attempts can count toward limits |
| Text is empty or incomplete | Open the file and try selecting text; a picture of a page may need OCR |
| The answer misses a chapter | Supply the relevant section separately and compare the result with the source |
| A quotation or page number looks wrong | Verify it in the original; ask the model to state uncertainty instead of inventing a location |
OpenAI's file types page and File Uploads FAQ are the sources for current upload rules. Uploading a document for analysis is different from opening it in a book reader.
What if the source is a photograph of a book?
A scanned page needs text recognition before it becomes editable text. On iPhone or iPad, EasyReadAI's paper-book OCR workflow lets you photograph or select pages, review the recognized words against the images, and save notes. Basic OCR is local; optional AI image proofreading sends image and text data online.
This creates reading notes, not an automatically reconstructed EPUB. Check important quotations before using them elsewhere.
Continue reading after the file question
If you want to read EPUB, PDF or TXT and collect notes on Mac, iPhone or iPad, EasyReadAI provides a book library and passage capture. Select text, capture it, type a question and tap Ask AI. Saved notes can contain the passage, your own comment and the AI reply, without leaving a permanent page highlight.
On iPhone and iPad, saved excerpts can also appear in a reading quote widget. Books and notes stay local to each device; AI requests are online and there is no automatic cross-device library sync.
Compare reading workflows · View EasyReadAI
Official upload documentation checked on September 20, 2026. This article does not claim an account-by-account EPUB compatibility test.