How Readers and AI Assistants Are Now Finding Books

Pick one of your backlist titles, then think of a well known book that sits close to it in subject or style.

Open an AI assistant and ask what the well known book is about, then ask for something to read next if you loved it. See whether your title appears in the answer. Then ask the assistant about your own title directly, check whether what it says is accurate and complete.

The results will show you where your title information is strong and where it needs attention.

How Readers Find Books Is Changing

Readers have long found books through bookshops, retailers, reviews, social media and search engines. Assistants such as ChatGPT, Gemini, Copilot and Claude now sit alongside those routes. Readers use them to ask for recommendations, compare titles and look up authors. Instead of a page of links they receive a conversational answer to their question.

For a publisher, this adds a new channel to the mix. It’s still early days, and the numbers are modest, but the habits are forming quickly, and it tends to suit exactly the kind of reader who buys a book on a recommendation.

Why AI Assistants Matter for Publishers

An assistant can only recommend what it can read and make sense of. When it draws on a publisher’s website, it is more likely to recommend titles where the details are complete, consistent and easy to interpret: what the book is about, who wrote it, who it suits, what format it comes in and where it can be bought.

Assistants read a page as text and data, so the details on it count for far more with them than the colours, imagery and mood around it.

That matters more as assistants begin to shop on a reader’s behalf. They compare formats, prices and availability against what the reader asked for, and may skip pages when they can’t work out the basics.

Publishers are well placed to make the most of this. Your catalogue already holds this information in detail, because it has been maintained for retailers and distributors for years. The opportunity is to make sure the same quality of information is visible on your own website, where readers and assistants alike can find it.

Four Questions to Ask About Your Own Catalogue

These four questions are a quick first look at where you stand. You can answer them without any specialist knowledge:

1. Is each title described fully? Pick a backlist title and ask whether a reader could tell from its page what the book is, who it is for and why it is worth reading.
2. Is the information consistent? Assistants gather what they know about a title from several places. If your site, your distributor and the retailer listings describe the same book differently, an assistant may repeat the wrong version or leave the title out.
3. Do your pages answer the questions a reader would ask? Questions like “is this suitable for a book club?” or “is there an ebook?” are the kind an assistant gets asked every day.
4. Do you know who is visiting? Your site analytics will show you whether assistants are already sending readers your way. Your server logs will show when the assistants themselves are visiting, including when they go looking for a way to read your catalogue directly.

If you are not sure about any of these, we can answer them for you by looking at your site the way an AI assistant does.

How AI Assistants Read Websites

An assistant doesn’t see your page the way a reader does. It reads the text and any labelled facts the page carries. Publishers understand metadata better than most, because ONIX has fed retailers and distributors for years. Your website needs its own version, called structured data, usually written as JSON-LD. It is a set of labelled facts coded into each page, such as a book’s title, author, ISBN, format, price and availability, in a form AI assistants and crawlers can read directly. Ideally those labels are generated from your ONIX feed. None of these labels are visible when you look at the page, so a site that looks great to a human can carry little of what matters to AI, or carry it incorrectly. The real test is to run a page through an AI readiness scanner, which shows what an assistant can actually read from it.

Measuring the Shift in AI Site Traffic

Visits referred by AI assistants are a small but growing share of total traffic for most publishers today. They’re worth tracking; a few months of data will show you which titles assistants are sending readers to, and whether those visits turn into clicks or sales on your site. Readers who arrive this way often come with a clear intention, as they’ve already been told your book fits what they’re looking for.

It also suits a long tail catalogue especially well. Assistants answer very specific requests, so a backlist title with a clear description can be the perfect answer to a question that no marketing campaign would ever have targeted.

Where This Trend Is Heading

Over the next year or two, it’s likely AI assistants will move from recommending books to buying them on a reader’s behalf, which makes a publisher’s own site, with accurate details and clear paths to purchase, an even more valuable asset.

Your site will still be read by people first, it should keep its character and its design. The aim is a page that feels right to a reader but also gives an assistant the facts it needs.

Publishers who treat this as an extension of the careful catalogue work they already do will be in a strong position to take advantage of AI assistants.

If you would like to know how your site reads to an AI assistant, we would be glad to take a look for you.