1 Your reference list
Search a title you have seen, score it 1 to 10, and put it where it belongs. The order is the ranking: #1 weighs most.
Nothing scored yet.
Use “Fill with 5 popular items” above for a demo list to try the score-and-rank idea; edit or clear it whenever you like.
The order is the ranking — #1 counts most. Use ▲▼ to move a title up or down the list.
Optional: rate genres (1-10)
Rate a genre in general, without adding a title. One 1-10 score per genre; 5 means “no opinion”. Rating thriller 10 and romance 1 leans the suggestions that way — also when no title of that genre is on your list.
2 Your top 10
Nothing here yet — press Show me 10 suggestions above and your ten picks land in this panel, each with a percentage and the reason for it.
⚠️ Estimates, not facts. They come from a script: no catalog holds everything, genres are broad labels, and taste is more than a list of scores. A hint, not advice.
3 To-watch list
Saved suggestions. Watched one? Search it in step 1 and score it there, so it starts counting in your taste profile. The ✕ takes it off this list.
Nothing saved yet — use “♡ To watch” on a suggestion.
Blocked titles (0) — never suggested again
Nothing blocked.
How the percentage is calculated (the default, no AI involved)
- Every title you score is split into features: genre, director/creator, decade, movie or TV show, country and language.
- Your 1-10 becomes one signed value: 1 → −1 (hate), 5 → 0 (no opinion), 10 → +1 (love). A title you dislike pushes a suggestion down, one you love pulls it up. The position is a second weight on top: #1 counts fully, the last entry half.
- Those features get a weight from your list — a genre that only shows up in titles you loved becomes a positive signal, one from titles you rated 0 negative. Shrinkage keeps a single title from dominating.
- Each catalog title is scored on how well it matches that profile, how similar it is to your loves (and how unlike your hates), plus a small popularity prior — squashed into a 0-100% estimate.
- The genre opinions and the preferences you scored each count as a few extra titles of that genre, director, decade, country or language, so both steer the next run.
- The final ten are picked with a mild diversity rule; blocked titles, to-watch titles (with that setting on) and anything your filters exclude are left out.
- Only the selected list is used, so a list of anime or of action movies gives suggestions for exactly that.
Optional: let an AI model re-rank the results
Off by default — the built-in script already produces the ten suggestions. Switch it on and a chat model receives your list and the thirty strongest candidates and rewrites the ranking. It needs an account, because every run is paid for by whoever asks, and your own API key from the provider you pick.
AI re-ranking is for signed-in accounts. Signing in also keeps your lists with your account instead of one browser.
The data, your account and your privacy have their own page: Data, accounts & privacy.