How Does AI Recommend Books and Movies?
Most recommendation engines match you to similar users. Here's how AI reads the content itself to understand why you love what you love, across every medium.
“People who liked X also liked Y.”
You’ve seen this everywhere: Amazon, Netflix, Spotify, Goodreads. It’s called collaborative filtering, and it’s how most recommendations work. It doesn’t understand why you liked X.
How traditional recommendations work
Collaborative filtering is simple and powerful:
- Find users with similar behavior to you
- See what they liked that you haven’t tried
- Recommend those items
It works because human taste clusters. People who read literary fiction often read certain authors, and people who watch Marvel movies watch other Marvel movies.
This approach treats your taste as a pattern of behaviors. It knows you watched these ten movies. It doesn’t know what drew you to them.
The limits of “people who liked…”
Consider two people who both loved The Shawshank Redemption.
Person A loves it for the prison escape narrative: the ingenuity, the patience, the triumph.
Person B loves it for the friendship between the two leads: the slow trust, the meaning found in connection.
Collaborative filtering treats them identically. They both rated it 5 stars, so they get the same recommendations, though they want different things.
This is why you’ve followed a recommendation and thought “I can see why someone would like this, but it’s not for me.” The algorithm found a behavioral match where you needed a taste match.
How AI recommendations work differently
AI-powered recommendations can read the content itself, not just track who consumed it. Where collaborative filtering asks what similar users consumed, an AI system asks what the properties of things you love are, and what else shares those properties.
The system can identify that you’re drawn to:
- Slow-burn character development
- Moral ambiguity
- Stories about people who don’t fit in
- Bittersweet endings
Then it finds items across any genre that share those properties.
Cross-media recommendations
Collaborative filtering is siloed. Amazon’s book recommendations don’t know what you watch on Netflix, and Spotify doesn’t know what you read.
An AI that understands your taste at the level of properties can work across media:
- “You love novels with unreliable narrators. Here’s a film that does something similar.”
- “Your favorite albums share a specific kind of melancholy. This book has the same feeling.”
- “Based on the themes in your TV shows, you’d probably like this anime.”
Your taste isn’t siloed by medium. You’re drawn to the same things whether they’re expressed in words, images, or sound, and AI can find those connections.
How Achriom uses AI for recommendations
In Achriom, the AI librarian knows your entire collection: books, movies, albums, TV shows.
When you ask for recommendations, it doesn’t just pattern-match to other users. It analyzes what you’ve collected and identifies the deeper properties:
- Themes that recur across your favorites
- Moods you’re drawn to
- Narrative structures you prefer
- Emotional registers you return to
Then it finds items that match those properties, even if they’re in different genres or media types than what you’ve added.
You can even ask specific questions:
- “What’s something like [favorite book] but different in [specific way]?”
- “I’m in [specific mood]. What fits that from my collection?”
- “What would I like that I haven’t discovered yet?”
Your librarian has the context to give useful answers, not generic recommendations.
What you get from it
Better recommendations save time. Instead of browsing endlessly or following suggestions that miss the mark, you find things you actually love.
The deeper benefit is understanding your own taste. When your librarian can tell you why you’re drawn to certain things, you learn something about yourself. Your collection becomes a mirror, not just a list.
Want AI-powered recommendations across your whole media collection? Try Achriom. It’s free to start.