Think about every AI chatbot window in your browser or sitting inside a dedicated app on your phone or desktop. They all want you to come to them, open their windows, drag files in them, paste text for context, before you even start asking them what you want. At WWDC 2026, Apple flipped this workflow on its head with the new ‘Siri AI.’
You can accuse Apple Intelligence of being late, but it looks like it’s worth the wait. The new “Siri AI,” powered by the next generation of Apple Intelligence, doesn’t really sit inside a sandboxed window.
From what Apple demonstrated during the WWDC 2026 keynote, the new Gemini and Apple foundational models-powered Siri AI lives at the system level. This allows it to reach into your files, messages, photos and apps to retrieve information based on the context of your prompt.
To me, that felt like the key structural difference highlighted in Siri AI, one that rival chatbots cannot replicate as of now (to the best of my knowledge).
The clearest demonstration of this ability of Siri AI came on macOS Golden Gate. The presenter selected three differently-formatted PDFs in any random folder, right-clicked them together and the new Siri field revealed itself in the pop-up menu, ready for a prompt. The presenter then asked Siri AI to compare them and it returned a structured table as part of its response.
But the real flex came when the presenter typed “Luke mentioned an electrical problem, which one will fix it?” and Siri searched across Messages and Mail, and added part of the private correspondence into the file comparison – while auto-spell checking the prompt query. No third-party AI app can read your OS context menu and your inbox in the same breath, so that was great to see.
Also read: WWDC 2026 recap: iOS 27, macOS Golden Gate, Siri AI and everything Apple announced
The same fusion runs through iOS 27 as well, according to another presenter’s demo during the WWDC 2026 keynote. “Show me photos from Shasta last weekend, add just the ones with Bryce, Madison and Quinn to our shared family album, and share with the whole family” executed a find-filter-add-share chain without even opening the Photos app!
Likewise, “give me directions to the arch with a stop at Jeff’s” had Siri identify a beach from an on-screen photo, dig Jeff’s address out of an old unsaved message, and build a multi-stop route – three capabilities, all in one sentence prompt. Haven’t seen that done before quite like that.
Privacy is the recurring justification, and Siri AI’s agentic ambitions go further in Passwords. In the demo example, while working with Safari, it was shown that Siri AI can navigate to each vulnerable site, sign in, and upgrade weak passwords to strong ones with a single tap. That’s next level orchestration.
Visual intelligence extends the reach beyond the screen, and even writing becomes contextual. Drafting in Mail or Messages, Siri AI mirrors how you actually communicate with a specific contact – manager at work versus friend – because it can see your real history with that person, claimed Apple
In all the above examples and demos, needless to say Siri is required to operate at the OS layer, drawing on the on-device semantic index, information from various apps, on-screen content and every app’s own action toolbox. That is what Apple Intelligence actually flexed overwhelmingly, during the WWDC 2026 keynote, confident in the knowledge that’s nigh on impossible for any standalone chatbot, however clever, to pull off right now.