Gemini Behind Siri and Pixel: How Google Won the Default

Google's Gemini is now the model layer under the default assistant on both major phone platforms. On Android, Gemini is replacing Google Assistant on phones, with the retirement starting September 4, 2026, according to a Google email reported by Chrome Unboxed, and it is the centerpiece of the Pixel 11 that went on sale August 20. On iPhone, Siri AI arrives with iOS 27 on September 14, running on Apple Foundation Models that Apple says were "custom-built in collaboration with Google and its Gemini models." No benchmark lead is worth as much as that position. What Google won is the model slot behind the default assistant on both major smartphone platforms, and on iPhone it won that slot without its name on the product.

Most model-war coverage asks whose model scores highest. This month shows why that is the wrong question for consumer AI. Assistants are distribution businesses, and whoever sits behind the power button or the wake word gets most of the queries by default. As of September 2026, Google holds that position on Pixel directly and on iPhone as a supplier. What follows covers what was actually agreed, what changes for users on day one, what developers should build for, and where rivals such as OpenAI and Anthropic still have room.

What exactly did Google win, and on which platform?

The two wins run through different channels, and the difference matters.

On Android, Google owns the product outright. Google announced the Pixel 11, Pixel 11 Pro and Pixel 11 Pro XL on August 12, with prices starting at $899, $1,099 and $1,299 and availability from August 20, according to the Google Keyword announcement. The Tensor G6 chip runs "the latest Gemini Nano model," and Google says it handles on-device AI tasks "up to 3.5 times faster while using up to 3.5 times less energy" than Tensor G5. Separately, Google is retiring the legacy Google Assistant on Android phones, tablets, paired Wear OS watches, earbuds and phone-projected Android Auto from September 4, 2026, according to Chrome Unboxed, citing an email Google sent to users. Once the update lands, users cannot switch back. The move had been delayed once: in December 2025, 9to5Google reported Google's statement that the upgrade "will now continue into 2026."

On iPhone, Google is the supplier, not the brand. On January 12, 2026, the two companies published a joint statement saying they had "entered into a multi-year collaboration under which the next generation of Apple Foundation Models will be based on Google's Gemini models and cloud technology," and that these models would power "a more personalized Siri coming this year." At WWDC on June 8, Apple's Apple Intelligence announcement confirmed that the new capabilities "are powered by the next generation of Apple Foundation Models, custom-built in collaboration with Google and its Gemini models." Apple's iPhone 18 Pro announcement sets the date: iOS 27 is a free update on Monday, September 14, and Siri AI rolls out "as a beta" that day for supported devices set to English.

	Android (Pixel 11 and other phones)	iPhone (iOS 27)
Assistant name	Gemini	Siri AI
Google's role	Owns the assistant, the model and the phone	Supplies the base models for Apple Foundation Models
Where requests run	Gemini Nano on device, plus Google cloud	On device and Apple's Private Cloud Compute
Key date	Pixel 11 on sale Aug 20; Assistant retirement from Sep 4 (per Chrome Unboxed)	iOS 27 and Siri AI beta on Sep 14
Google branding visible to users	Everywhere	None in Apple's product pages
EU availability	Varies by country and language, per Google	Siri AI not initially available on iOS in the EU

Why does Apple barely mention Gemini?

Read Apple's own pages and you have to look for Google. The June 8 Siri AI press release, which is Apple's main pitch for the product, never mentions Google or Gemini. The iOS 27 product page doesn't either. The credit shows up in a single sentence of the Apple Intelligence release and in the developer announcement. Apple's framing is that the joint statement promised "Apple Intelligence will continue to run on Apple devices and Private Cloud Compute."

That arrangement suits both companies, and it is the part most commentary misses. Apple keeps control of the user relationship, the privacy story and the brand. Google gets volume that its own Gemini app could not easily have won on iPhone, where Siri owns the side button. In practice that makes Google a component supplier, closer to a chip vendor than to a consumer brand. Nobody buys an iPhone for its modem, yet the modem supplier still gets paid on every unit.

The weakness in that position is also clear. A supplier relationship can be replaced. Apple has spent years building its own silicon to cut dependence on partners, and nothing in a "multi-year collaboration" suggests the arrangement is permanent. Google's lead on iPhone depends on Gemini staying ahead of what Apple can build or buy elsewhere. On Android, it depends on nothing but Google.

What changes for users on day one?

For iPhone owners the upgrade is less universal than "Siri gets Gemini" headlines suggest. According to Apple's footnotes, Siri AI in iOS 27 runs on "iPhone 16 models or later, iPhone 15 Pro, iPhone 15 Pro Max," among other devices. The most capable on-device features, including expressive voices and enhanced dictation, need an iPhone Air, iPhone 17 Pro or iPhone 17 Pro Max. iOS 27 itself runs on much older phones, back to the iPhone 11 and iPhone SE (2nd generation), so owners of a standard iPhone 15 or anything older can install the update but won't get Siri AI.

Three more limits apply at launch:

• It is a beta. Apple's own wording is that Siri AI "is rolling out with iOS 27 as a beta."
• English first. Apple says French, Japanese, Korean, Portuguese and Spanish support comes in October.
• Not in the EU on iPhone. Apple says it has delayed Siri AI in the EU for iOS 27 and iPadOS 27 over Digital Markets Act requirements. Craig Federighi said Apple does "not currently have a timeline" for EU availability on those platforms. Siri AI is available on macOS 27 and visionOS 27 in the EU.

For Android owners the change is less optional. When the Assistant retirement reaches a phone, Gemini answers "Hey Google" and the long-press of the power button, and there is no toggle back. Pixel 11 buyers get the most of it. Google's Pixel 11 feature roundup describes Gemini handling "multistep tasks across your favorite apps, expanding to more than 40 apps," along with a Gemini-powered voice input feature called Rambler and proactive cards that surface booking details and flight status. TechCrunch summed up the launch as few hardware changes and much more Gemini. Google says features are "available for select countries and languages to users 18+," and the Pro models include six months of Google AI Pro, which raises Gemini usage limits.

If you want Gemini's reasoning without either company's cloud, the local route is still open. We covered it in Running Gemma 4 E2B Offline on Mobile.

What should developers build for now?

The distribution shift gives developers a clearer target than they have had since assistants began. On both platforms the assistant is now the thing that calls your app, so the priority is making your app callable.

On iOS 27, Apple's developer announcement and its Siri AI announcement point to three levers:

1. App Intents. "Updates to the App Intents framework enable developers to connect their apps to Siri AI." If your core actions aren't exposed as intents, Siri AI can't act on them.
2. Spotlight. Apple says Siri AI's "personal context understanding extends to third-party apps when developers integrate with Spotlight." Indexing your content is now how Siri learns about it.
3. Foundation Models framework. Developers get a single Swift API to the Gemini-based Apple Foundation Models, through Apple rather than Google. The free tier is narrower than it sounds: Apple says developers in the App Store Small Business Program with fewer than 2 million total first-time App Store downloads can "access the next generation of Apple Foundation Models running on Private Cloud Compute at no cloud API cost."

Apple also says developers can "easily leverage models of their choice, like Claude and Gemini," or any provider that "implements the new language model protocol." That makes the model layer inside apps pluggable, even though Siri's own model is fixed.

On Android, the question is whether your app is one of the "more than 40" that Gemini's multistep tasks can drive on Pixel 11. Google has not published a general opt-in process in its launch posts, so for most developers the practical step is the agent tooling Google already ships. Our guide to Android CLI and Agent Mode is a starting point.

Doesn't the Google antitrust ruling stop this?

Not as written, and the wording explains how. On September 2, 2025, the court in United States v. Google issued its remedies decision. The Justice Department's summary says Google is "barred from entering or maintaining exclusive contracts relating to the distribution of Google Search, Chrome, Google Assistant, and the Gemini app," and cannot tie revenue-share payments to keeping those products on a device "for more than one year." The DOJ also said the remedies "will reach GenAI technologies and companies."

Our reading, which is analysis and not a legal opinion: the remedies target distribution of Google-branded apps. The Apple arrangement, as both companies describe it, is a model supply deal in which Apple builds its own Foundation Models on Gemini and ships them under the Siri brand. That is a different transaction from paying Apple to preinstall the Gemini app. Apple's parallel move to let developers plug in Claude, Gemini or any provider that implements its protocol also makes the platform look less exclusive. Competitors or the Justice Department may still test that distinction.

Who loses, and where is there still room?

OpenAI has the clearest loss. Apple's iPhone user guide has documented a ChatGPT extension for Apple Intelligence. But answering a question through an optional extension is a long way from being the model Siri itself runs on. OpenAI's route to iPhone users now runs mainly through its own app.

Anthropic gains a developer-level foothold rather than a consumer one. Apple named Claude as an example model in its developer frameworks, and Apple's developer announcement says its tools bring "today's best models and agents from Anthropic, Google, and OpenAI" into developer workflows. That fits Anthropic's strength with builders. It does not put Claude behind the side button.

Open-weight models remain the hedge for anyone uneasy about one company supplying both platforms. The open-source LLM landscape has matured enough that apps can run capable models locally, and Apple's language model protocol invites exactly that.

Google's own risk is concentration. When one model family sits under both default assistants, its mistakes, outages and policy choices reach almost every smartphone user at once. That is a reliability and regulatory exposure that no benchmark measures either.

What to watch next

• Siri AI's beta quality after September 14. If the new Siri ships rough, Apple takes the blame, not Google. That protects Google's brand but may pressure the partnership.
• The October language expansion and whether the EU standoff under the Digital Markets Act moves at all.
• How far the Assistant retirement has rolled out by the end of September, and whether users push back on losing the option to switch back.
• Terms of the Apple–Google deal. Neither company has published financial terms. Treat any figure you see as a report, not a disclosure.

The takeaway for operators: stop tracking which model tops a leaderboard and start tracking which model sits behind the default. As of September 2026 that is Gemini, directly on Android and underneath Apple's own models on iPhone, and the practical response is to expose your app to both assistants now, through App Intents and Spotlight on iOS and Gemini's agent integrations on Android.

Frequently Asked Questions

Does Siri now use Google Gemini?

Indirectly. Apple says Siri AI in iOS 27 is powered by the next generation of Apple Foundation Models, "custom-built in collaboration with Google and its Gemini models." Users see Siri, not Gemini, and the companies say Apple Intelligence continues to run on Apple devices and Private Cloud Compute. The January 2026 joint statement describes the models as "based on Google's Gemini models and cloud technology."

When is iOS 27 released, and which iPhones get Siri AI?

iOS 27 is a free update on Monday, September 14, 2026. Siri AI rolls out that day as a beta in English on iPhone 15 Pro, iPhone 15 Pro Max and all iPhone 16 models or later. The most capable on-device features need an iPhone Air or iPhone 17 Pro model. Other languages follow from October, and Siri AI is not initially available on iPhone in the EU.

Can I keep Google Assistant on my Android phone?

No. According to Chrome Unboxed, citing an email Google sent to users, Google began retiring Assistant on Android phones, tablets, paired Wear OS watches, earbuds and phone-projected Android Auto on September 4, 2026, rolling out over several weeks. After the update, Gemini replaces it and you cannot switch back. Cars with Google built-in keep Assistant for now.

Does the Pixel 11 cost more because of Gemini?

The Pixel 11 starts at $899, the Pixel 11 Pro at $1,099 and the Pixel 11 Pro XL at $1,299 as of September 2026, according to Google. Core Gemini features are included. The Pro models add six months of Google AI Pro, which raises Gemini usage limits.

Can iOS developers use Gemini or Claude inside their apps?

Yes. Apple's June 2026 developer announcement says developers can use the Gemini-based Apple Foundation Models on Private Cloud Compute, free of cloud API cost for App Store Small Business Program members with fewer than 2 million first-time downloads, or plug in "models of their choice, like Claude and Gemini," from any provider that implements Apple's new language model protocol. Siri's own model is not swappable.