What happened

On 14 September 2026, Apple began rolling out iOS 27 worldwide, with iPadOS, macOS and its other platforms updated the same day. The headline feature is an entirely new Siri, which Apple calls Siri AI.

Apple’s announcement explains where it comes from. Siri AI runs on the next generation of Apple Foundation Models, “custom-built in collaboration with Google and its Gemini models,” and those models run on the device and on Apple’s own Private Cloud Compute servers.

The deal behind it went public on 12 January, in a joint statement from Apple and Google. The two companies announced a multi-year collaboration under which Apple’s next-generation foundation models would be based on Google’s Gemini models and cloud technology. No price was disclosed. The widely quoted figure of about $1 billion a year traces back to a Bloomberg report last November citing people familiar with the matter. It isn’t an official number.

The rollout has limits worth reading closely. Siri AI is a beta, in English, with French, Japanese, Korean, Portuguese and Spanish coming in October. It won’t be available at first on iPhones in the EU, and it isn’t available in China for now. It needs an iPhone 15 Pro, or an iPhone 16 or later.

Why it matters to you

Nobody downloads Siri. It’s part of the operating system: hold the side button and it’s there, and the user never has to make a decision to go and ask an AI. For the past couple of years, brands have worried about whether they show up in chat windows. Now that question has moved into the phone itself.

What follows is our inference, not something Apple has said. If Siri AI’s models are built on Gemini, then much of what Gemini has read and retained about your brand is likely to carry through into Siri’s answers. Apple customised the models, so the two won’t be identical, but a model’s basic picture of the world doesn’t get rebuilt from scratch just because it’s been fitted into a new casing.

That inference raises the stakes on Gemini. Until now, Gemini was one of three engines to watch, alongside ChatGPT and Perplexity. Now whatever Gemini knows may have one more outlet, and it’s built into the iPhone. If you still haven’t asked ChatGPT how it describes you, you probably haven’t checked Gemini either.

There are two things Apple hasn’t addressed: which search provider Siri uses when it goes to the web, and whether its answers will show links or say where the information came from. The announcement says only that Siri AI will “leverage information from across the web.” If it answers out loud and without links, the user gets a single conclusion and no way to click through and check it.

What to do about it

Chinese isn’t on the October list, and Apple hasn’t given a timeline for Traditional Chinese or for Taiwan either. What you can prepare in the meantime is making sure Gemini, along with ChatGPT, which you should be watching anyway, gets you right in Chinese. There’s no “Siri optimisation” setting; the work goes into the content the models can read.

The difficulty is that you’re preparing for an outlet you can’t see yet. There’s no Chinese Siri to test, so how it will describe you, and whether it will cite anything, can only be inferred from what Gemini and ChatGPT say in Chinese today: how they describe you, which line is wrong, and roughly where they picked that line up. The part a model has memorised updates slowly, so a correction you make today will mostly have to wait for the next version of the model to be absorbed. And when it searches live, you can’t see where it’s looking, so you have to work backwards from the answer itself. It’s a long-cycle job, and it starts with listening carefully, line by line, to what they say in Chinese now.

Before Siri starts speaking Chinese, set up a time with us and see how ChatGPT and Gemini introduce you in Chinese today.

Further reading