ShelfGarden is a place to keep the books you want to remember. Most of us save book recommendations as screenshots and then never find them again. You upload those screenshots, ShelfGarden works out which books are in them, and you put the ones you want on shelves.
You don't have to read anything to use it. Collecting and arranging books is a perfectly good hobby on its own. Print, ebooks and audiobooks all count, and so do rereads.
Reading your screenshots. When you approve an import, each image is sent to an AI model, which reads it and lists the book titles and authors it thinks are there. We then look those up in ISBNdb, a book database, to get covers and details. The model gets things wrong sometimes, especially with blurry photos, partial titles or books that are only mentioned in passing. That's why every result goes to Review first and nothing is added until you accept it.
Recommendations (private testing). The Oracle is our planned premium recommendation feature, currently in private testing. It suggests real books — each one verified against ISBNdb — based on what you've loved. AI helps describe each book's qualities (its setting, themes, pacing and so on) using only the publisher's description; the AI has not read the book, and every quality carries a confidence level. Matching to your taste happens on our own systems using your ratings and preferences, and only if you turn on AI reading personalization in Settings. Your notes and reading history are never sent to the AI. We won't open it to everyone until we're happy with its accuracy.
We don't sell your data or use it for advertising. We're still confirming how long the AI provider keeps images it processes, so we don't make promises about that yet. The privacy policy has the details.
The illustrated garden, library scenes and library-corner furniture were made with AI image generation tools. The corner is assembled from separate painted pieces, with books and collection labels added in code. The logo and procedural book spines are drawn in code. Book covers come from ISBNdb. Your own notes, reviews and shelf names are written by you and are never generated.