Siri spent years seeming like it didn’t quite fit the level of quality Apple aspires to. Worse, the company announced capabilities it wasn’t able to release, and features that have disappointed. You can tell this wasn’t just users carping: more than one head of the Siri effort has left the company altogether.
Apple Intelligence was more of a success, if limited. The company, making use of mostly on-device computation, focused their use of Apple Intelligence on specific tasks within apps or systemwide. I think many of these attempts were boring or unnecessary: generating custom emojis, Image Playground for slightly icky picture generation, and the global Writing Tools, which were worse than many previous generations of non-AI-based assistants.
The reboot has come, however. With Siri AI and Apple Intelligence in iOS 27 and iPadOS 27, Apple is resetting our expectations for Siri and adding Apple Intelligence features in many more apps. Those features are a mix of machine learning and large language models (LLMs), and some tie into off-device, data-center-based LLM assistance.
Learning by rote versus guessing the next word
Machine learning typically refers to training an algorithm with a massive amount of data so that it can loosely pattern-match a result. In Photos, the loose searching you can do to find matches for people, places, buildings, and things relies on this. Likewise, voice transcription and Live Text are largely machine-learning-based.
LLMs, by contrast, use huge bodies of text and images to produce essentially sophisticated autocompletion. Given a set of words, what is the constellation of the next most likely one? Randomly selecting among those provides something that sounds like human writing. However, they are actually providing a statistical guess. Image generation is more complicated, but it relies on extracting features and then inserting them seamlessly, with the same degree of random guessing about which one to add.
Let Siri AI answer your questions
While Apple has rebranded Siri as “Siri AI” on supported devices and in their marketing, you won’t see that name on your devices. It’s still called Siri. But everything is affected by the availability of Apple Intelligence. Whenever you invoke Siri on an Apple Intelligence-supported device, you’re invoking Siri AI.
Instead of the glow around all four edges of your device in iOS 26/iPadOS 26, Siri in iOS 27/iPadOS 27 appears as an orb at the top of the screen. The design cleverly hides the non-screen areas in Dynamic Island on the many iPhone models with that carveout.
You can bring up Siri (by default), as in previous releases, by saying “Siri” or “Hey Siri,” or by pressing and holding the side button. However, the change introduced in iOS 26 that let you double-tap the bottom of the screen is gone. Instead, you can swipe down on a Home screen to reveal the new combination of Spotlight and Siri: a field labeled Search or Ask. You can type into that field, or tap the Dictation icon and talk to Siri that way.
Siri now provides much more accurate and detailed answers than before, even reaching into other apps to provide more information. Ask it a question, and it parses the results instead of trying to squeeze them into a format it understands.
Here are the most significant things Siri AI can do that its predecessor Siri could not:
- Use your data to produce better results or find things. One of the biggest ongoing complaints about Siri is that for software that ostensibly had access to our email, messages, contacts, calendar entries, photos, and other private data while running entirely on our devices, it sure didn’t know much about us. With Siri AI, you can ask, “Who did I have lunch with in 2013 in New Haven?” and it may be able to answer.
- Answer questions about what’s on screen. You could be looking at a webpage, reading a Wikipedia entry, having a conversation in Messages, or writing an email, and can ask Siri for more information about or to explain something that’s visible.
- Engage in a back-and-forth conversation. Siri has become something more like a chatbot, though it’s not quite as open-ended as the ones from Anthropic, Google, OpenAI, and others. However, it retains context and remembers aspects of conversations to provide better insight in the future.
- Create a plan. You might ask Siri to help you figure out the best sights to see in New York. Unlike chatbots that rely solely on synthesizing vast data stores, Siri can use your information, Apple’s human-made travel guides, map data, and other information to produce something that could be much more useful.
- Use a Siri app. You can revisit Siri conversations using the Siri app, the first time Apple has offered that kind of interaction with Siri.
Isn’t this invasive?
Didn’t we think one of the best things about Siri, and Apple’s overall approach to data privacy, is that everything stayed on our devices and within our control? Or, if it were synced in the cloud, we had two tiers we could choose from to protect it, including end-to-end encryption (E2EE) for nearly everything?
Yes, and that remains the case. Siri AI performs as much processing as possible on your device. When it needs to tap into more or different AI models, it uses what Apple calls Private Cloud Compute, which includes servers under Apple’s control and those that are part of Google Cloud. Apple says our “personal data is not stored nor made accessible to Apple or anyone else.” There are limits to that promise worth knowing.
Where Siri AI is available
- European Union: Siri AI will first appear in macOS and Apple Vision Pro. It’s unclear when they will roll out to iPhone and iPad.
- China: Siri AI and new features in Apple Intelligence were originally slated to be delayed at launch in China (PRC), with Apple saying regulators need to review changes. However, news reports in mid-August 2026 indicated Apple had trained an AI model with a Chinese partner that would allow it to offer Apple Intelligence features in China at launch.
