← Back to Book Carousel

How Book Carousel works

An AI that doesn’t chat. It listens.

Most apps that call themselves “AI” today send what you type to a huge language model in someone else’s data centre. Book Carousel doesn’t. Its recommendation AI was written from scratch for one job only: finding books that fit where you are right now.

4 min read · 28 September 2026

No chatbot, no language model

Book Carousel doesn’t use a large language model, and it doesn’t pass your answers on to an outside AI service. There’s no prompt, no generated text and no model trained on the whole internet.

The engine is a set of algorithms written specifically for Book Carousel. It runs as our own code on rented servers, and the only thing it consults is our own library of books.

What happens to your answers

How a recommendation comes together Your answers become a reading profile. Every book gets a score against it. The best matches are mixed into a list. Your thumbs up or down shift the profile, and the next list adapts. 1 Your answersOnly what you choose to share 2 Your reading profileNeeds · personality · mood · pace 3 Every book gets a scoreCompared with your profile 4 A mixed top listNot five books from one shelf 5 Your 👍 / 👎Shifts your profile right away LEARNS
Everything happens inside Book Carousel’s own engine.
  1. Your answers become a reading profile. What you need from a book right now (company, a way out, insight or simply a good story), plus your mood, how you like to read and a simple personality type. The personality part is translated into the Big Five traits psychologists use to describe people.
  2. Every book gets a score. The engine compares what each book offers with what you need. That match counts most: about half of the score. How well a book suits your personality, your mood, and the complexity and pace you’re up for makes up the rest.
  3. The list gets mixed, so you don’t end up with five books from the same shelf.
  4. You rate, it learns. Every 👍 or 👎 shifts your profile straight away, and your next list reflects it. Ratings are saved without your name or IP address, and over time they teach the engine which kinds of books really help in which situations.

So where’s the AI?

“AI” doesn’t have to mean a chatbot. Book Carousel uses classic machine learning: models that learn patterns from real ratings, and probability-based updates that adjust your profile every time you rate. It’s small, it can explain its choices, and it’s built for exactly one purpose.

Why that’s good for you

What it can’t do (yet)

Book Carousel is an independent project from Germany. Ideas, criticism, or a book it really should know? Write to us.

Curious what it picks for you?

About two minutes · no account needed

Find my books →