The metadata inside your ebook file
You filled in a form. Title, subtitle, author, description, categories, language, publication date. The store took it, built your listing from it, and that felt like the end of metadata.
It was half of it. Inside the EPUB you uploaded is a second, separate record of the same facts — and unlike the form, you probably never looked at it. It was written by whatever produced the file, it may say something you would not have typed, and it is the copy that goes wherever the file goes.
Two records, one book
The store's record lives in the store's database. It drives the listing, the search index, and everything a browsing reader sees. That is the one with its own guide: Metadata: the fields that decide who ever sees your book.
The file's record lives inside the EPUB, in the package document that every valid EPUB has to carry. It holds a short, fixed set of fields:
- Title — the book's title as the file states it.
- Author — the creator, with a role.
- Language — a language code, and it is required.
- Identifier — a unique string for this edition, required, and usually the ISBN.
- Publisher — the publisher of record.
- Publication date — when this edition was published.
- Accessibility declarations — a statement of what the file supports.
For the listing itself, the store's form usually wins. So the obvious conclusion is that the inside copy does not matter. It is wrong, and it is wrong in four places at once.
Where the inside copy is the only copy
A reading app's library. Open a sideloaded file on a phone or an e-reader and the shelf shows the title and author from inside the file. The store's record is not there. It never left the store.
A screen reader. The declared language is what tells assistive software how to pronounce the words. Get it wrong and an English novel is read aloud in French phonetics — not badly translated, mispronounced, sentence after sentence.
Library and catalog systems. Institutional systems read what files declare and match records on identifiers. A file that disagrees with the record attached to it is a file a system has to guess about — which matters the moment you enter that channel, as Getting your ebook into libraries describes.
Anywhere the file goes on its own. A review copy you email. A file a reader sideloads. A migration from one platform to another. Strip the store away and the metadata inside is all the file knows about itself.
What actually goes wrong
A title that is not the title. Files come out of conversion pipelines carrying the source document's name — "Document1", "novel-final-v7", "Untitled 3". The listing is fine. The reader's library shows "novel-final-v7" on the shelf, and that is the version of your book they look at every day.
An author who is not you. The creator field is frequently populated from the registered user of whatever word processor produced the manuscript. That might be your real name when you publish under a pen name, an old employer, a co-writer, or whoever installed the software on that laptop. It is a privacy problem as often as a tidiness problem.
A missing or wrong language. Two separate failures follow. The screen-reader one above, and a quieter one: hyphenation and line breaking are language-specific, so a reading app that justifies text will break words by the wrong rules or stop hyphenating altogether. A file with no language declared is invalid — see What makes an EPUB pass validation — but a file with a plausible, wrong one passes every check and misbehaves quietly.
An identifier that does not match. The identifier inside the file should be the identifier of this edition — the ebook's own ISBN if it has one, and not the paperback's. Pasting the print ISBN into the EPUB is a common slip, and it tells every system that reads it that this file is the paperback. Do you need your own ISBN? covers when you have one to put there at all, and Buying your own ISBNs the arithmetic of holding enough of them that every edition can have its own.
In the Interior Formatter. The EPUB 3 is built from your book's own fields rather than from whatever your word processor left behind, with the language declared, accessibility metadata and ARIA roles set, and chapters numbered — it validates with zero errors and zero warnings under epubcheck. Start your ebook → · about the Interior Formatter
Accessibility metadata is a field, not a virtue
This is the part that has changed fastest. Whether an ebook is accessible is one question. Whether the file says so is a separate one, and it is the one machines can read.
Accessibility metadata is a declaration inside the file of what it supports: that the text is readable, that the navigation works, how images are handled, what a reader with a print disability should expect. Stores have begun surfacing that declaration on the book's page, and accessibility rules in some regions attach to it. Check what applies to you where you sell rather than taking any summary, including this one, as the position in your market.
The practical consequence is blunt. A file that supports everything and declares nothing looks, to anything reading it automatically, exactly like a file that supports nothing. Accessible ebooks, and why they are now everyone's problem covers what the support itself involves.
Checking yours
You do not need a tool you have to learn. An EPUB is a zip archive, so a copy renamed to .zip and expanded gives you the package document as plain text you can open and read — the fields above are near the top of it. Reading it takes a minute. There are also free ebook library managers that show and edit the same fields directly.
Do it once for every file you have published. And if you fix something, remember the file already out there does not change retroactively: Updating an ebook after it is published covers what re-uploading reaches and what it does not.
Make your ebook
The tools are free to use: bring your manuscript once and get both a print PDF and an EPUB out of the same source, so the two editions cannot drift apart. You only pay when you want the files without the preview watermark.
Next: What makes an EPUB pass validation · More: Accessible ebooks, and why they are now everyone's problem · Metadata: the fields that decide who ever sees your book · What a store actually shows a browsing reader