How Shelf & Screen works
Your library stays on your device
When you choose an export file, JavaScript in this tab reads it, matches the titles and ranks picks. Nothing is uploaded: there is no server to upload to. The site is static files on GitHub Pages, and every page carries a Content-Security-Policy that stops it from connecting to any other site. To check, open your browser's developer tools on the Network tab while you import: the only requests are for this site's own files.
Your library is saved in this browser's local storage so it's here next time. Clear everything in the Library tab deletes it. A shared taste profile contains only the summary you see before you share it, and it travels in the part of the link after #, which browsers never send to a server.
How picks are ranked
- Your ratings, read relative to you. Each rating is compared with your own average and spread. A title you put on a favorites shelf or liked counts extra, a did-not-finish counts against, and a title you finished without rating counts a little for.
- A taste profile. Every catalog title carries tags for genre, themes, mood, style, setting and pace. For each tag we work out how you rate the titles that have it, and lean on it only as far as your library supports.
- Hand-written links. Each catalog title also lists a few titles its fans should try next, with the reason. If you rated a title highly, the titles it points to get a lift.
- Ranking and reasons. 1670 titles (825 books, 845 films) are scored against your profile. Near-duplicates are spread out, and every pick names the titles of yours behind it, the traits you rate highly, and any trait you tend to rate low.
The picks come from a fixed catalog that was written by an AI model ahead of time and checked by machine. No AI runs while you use the site, and nothing you import trains anything.
Where the titles come from
Every title, and every title it recommends, is checked each month against Open Library (books) and Wikidata (films). A title that can't be found there is left out. 2389 titles passed the latest check (2026-11-22). To recognize more of your library, we also match it against the most-read books on Open Library and the films most covered on Wikipedia. Those help us understand your taste but are never offered as picks. Both sources publish their data as CC0. Hooks, tags and reasons are original writing.
Corrections
Spotted a wrong year, a bad tag or a pick that makes no sense? Open an issue. The catalog is plain text in the repository.