Summary: Good metadata cannot stay locked inside the EPUB. How to get a book's description into the catalogue and onto public pages so AI search can find it.
In The reader of your book's metadata is now a machine we looked at the editorial decisions that need to be recorded in the files themselves: accessibility, text and data mining reservation, and how the work presents itself to the systems that handle it.
Now let us follow that information out of the EPUB.
Imagine your publishing house has just finished reviewing a digital book. The authorship is right, the subjects are well chosen, the accessibility features are described. Everything checked. Then you open that same book's page in a bookstore and find a two-line blurb, with no context, no indication of audience, and nothing at all about the accessible edition.
The work was done. Part of it was lost along the way.
This is where GEO, or Generative Engine Optimization, comes in. It sounds like the name of a Silicon Valley startup, but the idea is simple: improving how well your content shows up in the answers artificial intelligence tools give. For a publisher, one concrete route is making sure each book's good description reaches the places where the book can be discovered.
What changes with AI search?
In traditional search, a reader usually types a few words and browses the results. In a conversation with an AI, they can explain what they are after: "I want a short Brazilian novel, set in the countryside, about family relationships."
See the difference? The question brings together origin, length, setting and theme. A generic category such as "Brazilian literature" covers only part of that need.
The study that introduced the concept of GEO looks precisely at how to improve the visibility of content in the answers generative engines produce. For books, one practical application is offering information that lets each work be matched to readers' questions.
When a web search is involved, where that information is published matters too. Google explains that its AI search features may run several related searches to build an answer with supporting links. That gives a publisher a very concrete reason to look after its catalogue pages.
The description has to answer the reader's question
Imagine introducing an author to a bookseller. Giving only the name does not help much. Saying what they write, which readers tend to take to them, and what makes the new book particular opens quite a different conversation.
Something similar happens with a catalogue. Title, author and ISBN identify the work. The blurb, the subjects and the intended audience add the context that lets it be matched to a specific search.
Take an invented example. A novel is listed like this:
A moving story about love, loss and discovery.
Now look at a more informative description:
Set in rural Paraná in the 1980s, the novel follows three sisters who return to their childhood home after their father's death. Between memories and conflicts, they have to decide the fate of the property and face the choices that drove the family apart.
The second description offers setting, period, characters and conflict. It is still a blurb, but it gives both the reader and the discovery systems more to work with.
Good metadata improves how a catalogue is identified, organised and presented. In AI discovery, it is reasonable to expect that clearer, more specific information also helps relevant matches between books and questions. How large that effect is varies with the tool and the query, and a recommendation depends on factors beyond the catalogue record.
For the publisher the task is quite concrete: put a faithful, useful description of every work into circulation.
How to get that information moving
In the previous article we looked at the package document, where the EPUB's metadata lives. It remains part of this care. EPUB uses Dublin Core elements to record information such as title, author, language and subjects.
But a description kept inside a file that is for sale may not be available to a web search. It is worth following the whole route: what is in the EPUB, what the publisher sends to its partners, and what appears on the public pages.
Every point on that route deserves a check. An updated blurb in the file does not mean the bookstore has updated its page. And a complete record at the distributor can show up abbreviated in the sales channel.
ONIX: the information that circulates in the trade
ONIX for Books standardises how information about books is communicated between participants in the publishing supply chain. The Inclusive Publishing metadata guide, with a contribution from the executive director of EDItEUR, explains that role.
For the publisher the practical question is this: what are we sending to the distributor, and what is reaching the bookstores?
Check the description, the subjects, the audience, the edition and the format. If the book belongs to a series, that relationship needs to be clear too. It is worth asking your distribution partner to review the record and checking how the data appears in the sales channels.
Information that is correct in the publisher's spreadsheet still has to reach the place where the reader will find it.
Schema.org: the information on the publisher's site
Schema.org has a type called Book, with properties for describing books on web pages: title, author, ISBN, language and format.
Think of that markup as tidy labels travelling with the page. The visitor reads the book's presentation; the systems that use structured data also find the identified fields.
Ask whoever looks after the site to assess this implementation. The markup should match what is visible on the page. Google recommends that consistency and makes clear that no special Schema.org is required to appear in its AI features.
The starting point is still a good page: complete information, readable text, and content that search engines can reach.
Accessibility has to be visible before the purchase too
If the previous article was about declaring accessibility inside the book, the next step is getting that information to the person choosing what to read.
Does your EPUB have alternative text on its images? Does it allow structured navigation? Are there accessibility limitations the reader ought to know about before buying?
The EPUB Accessibility 1.1 specification defines metadata for describing accessibility features, access modes and possible hazards. Book Industry Communication explains how ONIX allows that information to be shared along the supply chain.
Think of someone looking for an edition that has image descriptions. If that feature is recorded only inside the file, how are they to know while comparing the options in the bookstore?
Publishing that information helps the reader choose. It also gives a useful signal to discovery services that take the requirement into account. So check that the edition's real features appear both in the metadata and in how it is presented to the public.
I have 200 titles. Where do I start?
Pick ten books that matter to the catalogue. It can be a mix of new releases, steady sellers and works that deserve to be talked about again.
For each one, run through this review:
- Read the blurb as someone who has never seen the book. Does it explain the content, or could it serve for dozens of other works? Keep the invitation to read and add what makes this particular title what it is.
- Check the subjects and the audience. Use categories that fit and words that are specific. Include themes that genuinely have a presence in the work.
- Compare the channels. Open the publisher's site and two sales pages. Compare what appears there with the digital file and the distributor's record. Do they all identify the book, the edition and the format correctly? Has the most recent description reached those places?
- Look after the book's page. Bring together the blurb, the authorship, the technical details, the formats and the accessibility information. Review the structured data with the site team.
- Watch what the AIs answer. Try the sort of questions readers would plausibly ask and check the titles, descriptions and sources that come back. Note the tool and the date: answers vary, and a single query does not measure the result of the work.
If something wrong turns up, track down where it came from. Is the old blurb still on the site? Has a bookstore kept the record of another edition? Is there data that fell by the wayside?
Agree as well on who keeps the reference information inside the publishing house and who follows its update at the partners. That stops the next reprint, a new cover or a revised blurb from leaving the catalogue showing conflicting versions again.
At Booknando, looking after the file is part of producing the digital book. Talking to the people who look after the catalogue, the distribution and the website is what carries that work through to how the book is presented to the public.
Once you have looked at what the EPUB declares, it is worth looking at what the world can actually discover about it. Pick a title and follow that path. You may find a good description waiting inside the publishing house, ready at last to reach the reader.
