Siri AI Runs on Gemini: What It Does to Search Traffic

Siri AI now runs on Google Gemini — and that single architectural decision, unveiled at Apple's WWDC 2026 keynote on June 8, 2026, quietly reshapes how a reported 1.5 billion daily Siri requests get answered. For the first time, the same Gemini model family that powers Google Search and Android also answers questions on the iPhone, before a browser ever opens.

What's at stake for brands is visibility, not just rankings. Google AI Mode already established a pattern where an AI answer resolves the query without a click. Siri AI extends that pattern onto Apple devices and pushes it one step earlier in the journey — the answer can arrive the moment someone speaks, not when they reach a results page. A brand absent from the AI's answer can now be invisible across Google Search and Apple devices at the same time.

This guide covers what Apple actually shipped, why we read it as a second AI answer layer rather than a feature update, where Gemini does and doesn't answer inside Siri's tiered architecture, and a concrete plan for staying cited across both surfaces. Everything below is sourced from Apple's own announcement, the Google–Apple joint statement, and the search-industry reporting cited throughout — with vendor-stated and reported figures clearly labelled.

At WWDC 2026, Apple introduced Siri AI — described by the company as an entirely new version of Siri — and confirmed it runs on Google's Gemini models and cloud technology. The two companies first announced their multi-year AI collaboration in a joint statement on January 12, 2026, stating that the next generation of Apple Foundation Models would be based on Google's Gemini models. Bloomberg reported the arrangement is worth roughly $1 billion per year, though neither Apple nor Google has officially confirmed the financial figure.

Underneath Siri AI is a custom, Apple-tuned variant of Google's Gemini 3-family models — reportedly around 1.2 trillion parameters, though that figure does not appear in Apple's or Google's official WWDC materials and has not been independently benchmarked. What is confirmed is the surface area: Siri AI ships as a standalone dedicated app on iPhone, iPad, and Mac, with iCloud-synced conversation history, multi-step command execution across apps, on-screen question answering, and web access for current information.

The reach is the headline. Siri processes a reported ~1.5 billion user requests a day across an Apple install base of roughly 2.5 billion active devices — a figure that originates from Apple marketing, so treat it as vendor-stated rather than independently verified. Either way, it places a Gemini-backed answer engine in front of one of the largest consumer audiences in technology.

Here is the original argument of this piece: Siri AI is not a feature update, it is a second AI answer layer — and it sits earlier in the user journey than the first one. Google AI Mode was the first layer that resolved a query without a click to a website. Siri AI is the second, and it can answer the moment a user speaks, before they ever open Safari or type into a search box.

Consider the flow. Someone asks Siri "what's a good moving company in Cape Town?" If that query is answered in-line, the user never sees a Google results page, never scrolls past AI Overviews, and never clicks a blue link. The discovery moment happened on the lock screen. A brand that has invested years in Google rankings can be entirely absent from that answer — and the same is true if it's absent from the underlying Gemini model that now feeds both Google's answers and Siri's.

The compounding effect is what makes this strategically important. Because Gemini is the model family behind Google Search, Android, and now Siri, a brand that fails to earn citations inside Gemini-generated answers loses visibility across all three at once. The same omission that costs you a Google AI Overview citation can cost you the Siri answer too. That is a single point of failure dressed up as three separate channels.

One industry observer framed the commercial logic of the deal plainly for Google: as Matt Beucler, founder of Plura AI, put it, "This partnership lets Google capture value from it even without owning the device." For marketers, the mirror image of that statement is the warning — the audience you reach through Apple devices is now partly mediated by a Google model you have to optimize for the same way you optimize for Search.

Not every Siri request reaches Gemini, and that nuance changes how you optimize. Apple described a three-tier privacy architecture for Siri AI. Tier 1 (simple tasks) runs on-device using Apple's own models. Tier 2 (moderate complexity) routes to Apple's Private Cloud Compute. Only Tier 3 (heavy reasoning) routes to Google Cloud and the Gemini model. Apple's Craig Federighi said data is only used to execute the request and that the contract bars Google from training future models on Apple user data.

The practical reading is that simple factual lookups — a quick brand search, a local recommendation, a unit conversion — likely fall into Tier 1 or Tier 2, where Apple's own models answer, not Gemini. Heavier, multi-step reasoning is what tends to reach Gemini. That creates what we'd call a Gemini query floor: the GEO work you do for Google does not transfer cleanly to the simplest Siri queries, because those may never touch Gemini at all. Optimizing for Apple's on-device understanding is a different exercise from optimizing for Gemini's generative answers.

Most coverage treats Gemini as the only answer layer on the iPhone. The reality is a stack of surfaces a single query can hit before — or instead of — reaching a traditional results page. The table below maps that stack so you can audit where your brand is exposed. It is our own synthesis, drawn from the cited sources; treat the bypass-rate column as directional rather than a precise per-surface measurement.

The pattern that emerges from the stack is that two of the surfaces a query hits first are now Gemini-backed and zero-click by design. The traditional SERP — where classic SEO lives — has moved down the stack. Commercial and e-commerce queries still reach it (AI Overview coverage on e-commerce queries is only around 4%, versus 99.9% on how-to queries), but informational discovery increasingly resolves above it.

Siri AI does not arrive everywhere at once, and the gaps are a strategic variable. Apple confirmed Siri AI launches initially in US English only, and is not available in the EU or China at launch due to regulatory requirements — EU users get the Mac, Watch, and Vision Pro variant only, and China availability is under regulatory review. General availability is fall 2026, with a developer beta from June 8 and a public beta the following month behind a waitlist.

That uneven rollout means urgency depends on where your customers are. English-language brands in the US, UK, and Australia face near-term exposure; EU-based brands have a buffer window to get Gemini-ready before Siri AI reaches their market. The matrix below frames the launch picture so you can set the right priority by geography.

A Gemini-Siri world splits optimization into related but distinct disciplines, and treating them as one thing is how brands end up optimizing for a surface their customers don't actually use. Generative Engine Optimization (GEO) targets citations inside AI-generated answers — the working summary across the industry is that SEO gets you ranked, while GEO gets you cited. Personal Context Optimization (PCO) is an emerging, SEO-community framing (not official Apple terminology) for structuring first-party data, app metadata, and semantic markup so on-device models surface your brand. Classic SEO still owns the link layer for commercial and navigational queries.

The honest synthesis is that these are not competing choices — they are layers of one program. A brand that ranks well (SEO), is cited inside AI answers (GEO), structures its first-party data cleanly (PCO), and shows up across multiple models has coverage at every point a query can be resolved. If you're building that program from scratch, our agentic SEO and GEO engagements are built around exactly this layered model rather than chasing a single ranking.

Siri AI doesn't create the zero-click problem — it accelerates a trend that was already well underway. Zero-click searches account for 58.5% of all US Google searches (SparkToro / Datos), and Google's AI Overviews reduced clicks by 38% in a randomized field experiment run by Ahrefs in February 2026. AI Overviews now appear on roughly 48% of all Google queries as of March 2026, up from 34.5% in December 2025. Siri AI adds a second answer layer on top of all of that.

The other half of the story is the upside of being inside the answer rather than below it. Google AI Mode crossed roughly 100 million monthly active users in early 2026 with a 93% zero-click rate, and has been reported to scale toward 1 billion users globally — though that larger figure appears in secondary reporting and should be treated as reported until independently corroborated. Across that surface, citation is the new ranking: our own research found that brands cited inside AI Overviews earned about 35% more organic clicks and 91% more paid clicks than non-cited competitors on the same queries.

None of this requires a new toolset — it requires reprioritizing toward citation and structure ahead of the fall 2026 general availability. The work that earns a Gemini citation in Google's answers is the same work that earns one inside Siri AI's Tier 3. Start there, then layer on the Apple-specific and multi-model pieces.

For most teams the sequencing is straightforward: lock down classic SEO and structured data first (it feeds every layer), build GEO into your content process second, then add Apple-specific structuring and multi-model tracking. If your customers are in the EU, you have a buffer window to get all of this in place before Siri AI arrives — use it. You can monitor where you currently stand with the AI visibility tracking tools that monitor brand citations across Gemini and ChatGPT, and ground the citation work in the tactics from our GEO guide for getting cited inside AI-generated answers. For the device-side context behind Siri AI, see our breakdown of Gemini-powered Siri's context and on-screen AI capabilities and the iOS 27 Extensions that let users set Claude, ChatGPT, or Gemini as the default.