Apple TV's Search Bar Now Learns What You Like to Watch

Key points

• Apple ML Research published a system that personalises Apple TV search results after every single keystroke, even with just one or two characters typed.
• The system combines two separate AI models, one that reads the meaning of words and one that learns from what millions of users actually watch.
• It works across multiple languages, making it useful for Apple TV's global audience.
• The goal is to surface the right show or film faster, before the user has finished typing the title.

You type the letter "S" into the Apple TV search bar. The app has maybe half a second to guess whether you want Severance, Slow Horses, or a Spanish thriller you watched last month. That is a genuinely hard problem, and Apple ML Research has published details of how they are trying to solve it.

The challenge is called incremental search, which just means the app has to rank results after every single keystroke, not just when you finish typing. One or two characters give almost no clue about intent, so the system needs extra help from what it already knows about you.

Apple's approach uses two AI models running side by side.

The first, called TextEmb, is a multilingual text encoder, essentially software trained to understand the meaning of words and titles across different languages. It was fine-tuned, meaning given extra specialised training, on data about which shows people tend to watch together. If fans of one thriller often watch another, the model learns those two titles are related in meaning, even if the words look nothing alike.

The second model, called IdEmb, skips words entirely. It treats every show and film as a unique ID number and learns patterns purely from what people click and watch. Think of it as the "people who watched this also watched" logic, but baked into the search ranking itself.

At the moment you search, the system builds a quick profile of your tastes from your recent viewing history and blends both signals together to reorder the results.

Does this mean Apple is tracking everything I watch?

Yes, to a degree, though Apple has long said it processes personalisation data on-device where possible. The published research does not spell out exactly where this computation happens, so if on-device privacy matters to you, it is worth checking Apple's privacy settings under your Apple TV account. The paper focuses on ranking quality rather than data handling specifics.

In plain terms: the system is reading your watch history to make search faster. That is the trade-off.

For everyday use, the practical upside is real. If you half-remember a title and type a three-letter guess, a personalised system has a much better shot at surfacing the right show than a generic one that just matches letters. Apple TV carries a large multilingual catalogue, so the cross-language training matters too. Someone who watches content in both English and Italian should see relevant results regardless of which language the title is stored in.

Apple has not announced a public release date for this specific system, so it may already be running quietly in the background or still be moving through testing.