Which of your books does AI recommend?
Readers ask ChatGPT, Claude and Gemini what to read next. We check your list the way they do: which titles get named, which get passed over, and which changes are worth making first. It starts with a small pilot and a fixed quote.
Where is the AI shelf still open?
Each tile is a genre. The number is how often the top book is named when readers ask AI. A low number means no book owns that shelf yet. Tap a tile.
Open the full chart for every genre · AI Shelf, week 46: 2,304 real answers from ChatGPT, Claude and Gemini across 32 genres, UK panel. Updated every week.
How many titles are on your list?
Move the slider. This is the real work behind each check, which is why a pilot starts small.
A quick check asks 3 reader questions of 3 assistants. A full report asks 12 questions, twice each, of all 3. The web sources are Goodreads, Open Library, Wikidata, Wikipedia, YouTube and Reddit.
Four steps, agreed in writing first
Agree the list
Titles, market, genres, checks and budget are agreed before work starts. We use your catalogue to confirm each book and flag unclear editions.
Take a baseline
We ask the same reader questions for every title and record what the assistants say today, along with your Amazon pages and the web around them.
Get an action register
For each title: what we saw, the change we propose, who owns it and how to check it worked. Changes your team can actually make.
Measure the change
Log the changes, then ask the same questions again. Where data allows, we agree comparison groups and measure sales separately.
Bulk catalogue work is a managed pilot. Self-service catalogue upload is not available yet. No sales uplift or AI placement is promised. Read the method and its limits · See a real single-book report
Tell us roughly how many titles
We reply within one working day, with questions or a fixed quote.
- Title count, outputs, timetable, data handling and price agreed in writing
- We read public pages only and never need your Amazon or retailer logins
- We report uncertainty and model changes alongside the results
Publisher questions
Can we upload our whole catalogue ourselves?
Not yet. Bulk catalogue work is run by us as a managed pilot, so every book is matched to the right edition. Self-service upload is not available yet.
How many titles should a pilot have?
Enough to learn from and small enough to act on. Large catalogues are scoped in stages. Tell us the size of the list and we will suggest a first segment.
Do you promise more sales?
No. A score is a diagnosis. Public AI citations show what a model used. They are not a guarantee that a placement will improve sales. We measure before and after and report what changed.
What does it cost?
The price depends on the number of titles and checks. It is agreed in writing, with the outputs and timetable, before any work starts.
Which assistants and markets do you check?
ChatGPT, Claude and Gemini, for books listed on amazon.co.uk and amazon.com.