LLM search using Core Spotlight

A Podkey summary of LLM search using Core Spotlight, from WWDC 2026.Today is really about one idea: Apple’s Spotlight index is turning into something a language model can use directly, instead of just something a human searches through. The interesting part isn’t only that the model can query your app’s data, it’s how developers can shape that interaction so it stays accurate, efficient, and useful on real devices with real limits. There’s also a practical layer here around evaluation, UI updates, and filling in missing metadata, which is where this goes from neat demo to something you could actually ship.The basic integrationHow the tool-calling loop worksWhy metadata quality mattersKeeping the model in boundsPipeline search and custom stagesResolving who a person isHandling big result sets in the UIGetting full items back when neededMeasuring whether it actually worksWhy this feels meaningfulThis podcast was created with Podkey. Make your own at https://podkey.fm