Last reviewed by Robert Prime, 17 November 2026. Figures cover checks finished between 3 October and 14 November 2026.
Short answer: we looked at 104 Amazon book listings that authors checked with Book Score. AI assistants never named 91 of them (88%) when readers asked for a book like it. 101 of 103 (98%) had a title and subtitle that missed the words readers search with. Most of the fixes are free and sit with the author.
Who these books are: authors chose to check their own book, often because it wasn’t selling. So this is not a random sample of Amazon, and the figures will run worse than Amazon as a whole. We say so here rather than in a footnote.
The numbers
Our scoring rules changed during the period, and some checks didn’t apply to every book. So each line has its own count. Quote the line with its count.
| What we found | Books | Share |
|---|---|---|
| AI assistants never named the book in any answer to readers | 91 of 104 | 88% |
| Score band: weak / fair / strong | 80 / 19 / 5 of 104 | 77% / 18% / 5% |
| Title and subtitle miss the words readers use (some or all) | 101 of 103 | 98% |
| No Wikidata or Wikipedia entry | 87 of 104 | 84% |
| No audiobook edition | 81 of 103 | 79% |
| Few or no YouTube mentions | 80 of 104 | 77% |
| Not on the sites AI reads most for the genre (some or all) | 69 of 69 checked | 100% |
| Weak or missing category ranks | 73 of 103 | 71% |
| No hardcover edition | 66 of 103 | 64% |
| No A+ content | 53 of 98 | 54% |
| No A+ content, or only a thin one | 60 of 98 | 61% |
| The opening lines of the description need work | 41 of 101 | 41% |
| No editorial reviews section | 25 of 58 checked | 43% |
| Kindle price outside the 70% royalty band | 9 of 34 checked | 26% |
| No paperback edition | 16 of 103 | 16% |
| No Kindle edition | 15 of 103 | 15% |
| Cover weak at thumbnail size | 5 of 103 | 5% |
What it means, finding by finding
Most books are invisible to AI
For 88% of books, the AI assistants never named the book when Book Score asked them the kinds of questions readers ask. Only 5 of 104 books scored in the strong band.
AI assistants don’t read Amazon to decide what to recommend. They lean on reference pages, review sites, best-of lists and video. That is why the next few findings matter: they are the places AI learns about books, and most of these books were missing from them. Our guide to getting ChatGPT to recommend your book covers the steps.
The title and subtitle miss what readers type
This was the most common fault: 101 of 103 listings. A title can be creative. The subtitle is where most authors can add the words readers actually search for, such as the subject, the genre or the problem the book solves. Our book title guide explains how to choose them.
No reference entry
87 of 104 books (84%) had no Wikidata or Wikipedia entry. Reference pages are one of the places AI assistants learn about books and authors.
Missing editions
- No audiobook: 81 of 103 (79%). See audiobooks through ACX for UK authors.
- No hardcover: 66 of 103 (64%). See KDP hardcover.
- No paperback: 16 of 103 (16%). No Kindle edition: 15 of 103 (15%).
Each missing format is a reader who can’t buy the book in the form they want.
Few YouTube mentions, and absent from the genre’s key sites
80 of 104 books (77%) had few or no YouTube mentions. Every one of the 69 books we checked for it was missing from some or all of the sites AI cites most for its genre. We publish those sites, genre by genre, in the sites AI reads for book recommendations.
The Amazon page itself
- Category ranks: 73 of 103 (71%) had weak or missing category ranks. Our Amazon book categories guide explains how to choose categories you can rank in.
- A+ content: 53 of 98 (54%) had none, and 60 of 98 (61%) had none or only a thin one. See A+ content modules for books.
- Description: in 41 of 101 (41%), the opening lines needed work. Those lines are what a shopper sees before clicking “Read more”. See how to write a book description.
- Editorial reviews: 25 of the 58 checked (43%) had no editorial reviews section. See editorial reviews and author blurbs.
- Kindle price: 9 of the 34 checked (26%) were priced outside the band that earns the 70% royalty, so the author keeps less per sale. See the KDP 70% royalty.
- Cover: only 5 of 103 (5%) had a cover that was weak at thumbnail size. Covers were the one area most authors had right.
What to do first
- Rewrite the subtitle to include the words readers search for. It’s free, and it was the most common fault.
- Fix the first two lines of the description, so a shopper knows what the book is before clicking “Read more”.
- Add a Wikidata entry for the book and the author.
- Get onto your genre’s sites: the review sites and lists AI cites for your kind of book.
- Add the missing formats, starting with whichever your readers ask for.
To see which of these apply to your book, run Book Score on it. It’s free and takes about a minute.
How we counted
- Source: finished Book Score checks between 3 October and 14 November 2026: 113 checks from 83 different addresses, covering 104 different listings. Where a listing was checked more than once, only the latest check counts.
- Stores: 51 listings on Amazon UK, 46 on Amazon US and 7 on other Amazon stores.
- Left out: checks we know are ours: staff, test and repeat checks, and checks from our own servers and studio.
- Counts differ by line because our scoring rules changed during the period, and some checks only apply to some books. For example, the Kindle price check only applies to books with a Kindle edition we could price.
- Not a random sample: these are authors who chose to check their own book.
- No book is named. We don’t publish titles, authors, ASINs, prices or text from any listing.
- Next update: we will repeat this count every quarter, with the same rules.
Frequently asked questions
How many books were in the Book Score findings?
104 different Amazon book listings, from 113 finished checks run between 3 October and 14 November 2026. Where a listing was checked more than once, only the latest check counts. 51 were on Amazon UK, 46 on Amazon US and 7 on other Amazon stores.
How often did AI assistants name the books?
Rarely. For 91 of the 104 books (88%), the AI assistants Book Score asks never named the book in any answer to the questions readers ask.
What was the most common listing problem?
The title and subtitle. In 101 of 103 listings (98%), they missed some or all of the words readers use when they search for that kind of book.
Is this a random sample of Amazon books?
No. These are authors who chose to check their own book, often because it was not selling. Treat the figures as a picture of books whose authors were worried, not of Amazon as a whole.
How can I check my own book?
Run the free Book Score check at bookscore.publishing.co.uk. It reads your Amazon page and asks AI assistants the questions readers ask, in about a minute.
