Siri’s gemini upgrade still isn’t superhuman
The gist
Apple has officially made Google’s Gemini the brains behind Siri, but the new AI still can’t plan your life like a real assistant.
What to know
- In 2026, Apple locked in a multi-year, $1 billion-a-year deal with Google to power Siri using Gemini and Google Cloud.
- The Gemini-powered Siri now excels at quick, personal tasks—like finding TV schedules or retrieving old door codes—with impressive speed and context.
- Despite the upgrade, Siri still can’t reliably handle long-term planning, multi-app tasks, or ambiguous requests—so your digital assistant isn’t superhuman just yet.
Apple Bets Big on Google
Apple’s billion-dollar, multi-year pact with Google marks a historic shift—Siri’s intelligence is now fundamentally built on Gemini, signaling Apple’s first full embrace of an external AI core.
By spring 2026, Apple’s move from testing outside models to formally building Siri around one had become explicit. TechCheck reported on April 1 that Apple was “partnering with Google to power its artificial intelligence features, including a major Siri upgrade,” and that the arrangement was a “multi year partnership” built on Gemini and Google cloud technology, while also noting, “With Apple reportedly paying $1 billion a year for Google's AI…,” a sign this was a structured commitment rather than an experimental collaboration.
That made the May-to-June 2026 window less a rumor cycle than the moment Apple turned a strategic direction into a public product stance. More or Less had already summarized the outcome bluntly — “Google and Apple are doing a Siri deal” and “Siri is going to be powered by Gemini essentially” — so when Apple presented the new Siri at WWDC 2026, the announcement landed as confirmation that the company had formally embraced a third-party foundation model at the center of its assistant.
Siri’s Smarts, Still Limited
Siri now nails quick, context-rich tasks using privacy-aware AI, but still falls short of handling complex, multi-step requests or acting as a truly autonomous assistant.
The clearest evidence that Siri has improved is in ordinary, bounded tasks that used to require manual digging. On The Talk Show, a tester asked Siri AI for “the next two episodes of Magnum PI” and got an answer “within 30 seconds,” with Siri having “formulate[d] a spotlight search” to pull the most recent plan from messages; the same system could answer “what did Tola tell me his door code is?” by surfacing a weeks-old text, which is exactly the kind of privacy-preserving personal context Apple has been trying to make useful.
But those wins do not add up to a fully general agent. The successes described on The Talk Show are immediate, bounded actions rather than open-ended planning, and multiple commentators note that Siri still lacks broad access to the most sensitive data, dependable action-taking across apps, and the autonomy needed for ambiguous multi-step work; even a favorable November review concluded, “Siri AI is fantastic for 90% of use cases,” before stressing that it still does not replicate a real-world assistant that can carry long-horizon tasks on its own.
