Ten years later, Siri finally emerged from the system's overlay.

At WWDC on June 8th, Apple previewed its next-generation operating systems, including iOS 27 and macOS 27. Almost all the software updates pointed to the same thing: AI is transitioning from an accessibility feature to a system-level interaction interface. The most noticeable change is that Siri has transformed from an instantaneous voice-activated pop-up into a dedicated app with its own icon, supporting iCloud synchronization of conversation history. For the first time, its form resembles that of a standalone AI application like ChatGPT.

The previous generation of Apple Intelligence limited Siri to passive responses at the system level. It could help you polish text and summarize notifications, but it couldn't handle cross-app tasks, nor could it remember what you asked last time. This time, Apple is trying to upgrade Siri from an assistant that "understands you" to a proxy that "can do things for you." The impact of this AI update goes far beyond Siri itself. From refactoring and editing photos in the Photos app, to Safari automatically monitoring product price drops, to password management apps logging into websites and changing weak passwords for you, new AI capabilities are almost everywhere in the system.

Siri now has its own dedicated entry point, and conversations can finally remember the context.

For Siri to become a standalone app, the first problem to solve is "where do users find AI?"

In the past, Siri's entry points were scattered across voice activation, side buttons, and long presses on the lock screen, lacking a fixed visual anchor. After ChatGPT and Gemini occupied users' desktops as standalone apps, Siri's intangible state actually became a disadvantage: users didn't know what it could do, and they weren't used to repeatedly opening a tool without a user interface.

With a standalone app and iCloud syncing of conversation history, the experience is completely different. A user can have Siri analyze key terms of a contract on their iPhone, then follow up with further details on their Mac at lunchtime, and Siri can remember the context. This builds a long-term conversation asset, making Siri more like "an AI capable of continuous dialogue" rather than "a voice tool that answers single questions."

The underlying driving engine has also changed. Wired and several other tech media outlets confirmed at the event that Apple has launched the third generation of Apple Foundation Models and entered into a deep, multi-year partnership with Google Gemini to customize and develop the next generation of models using Gemini technology. At WWDC 2024, Apple's partner was OpenAI, and Siri could call ChatGPT in specific scenarios. Two years later, Google was added to the partner list. Apple is no longer relying solely on a single third party for its foundation models, choosing a more flexible, multi-party customization path.

From "Let me check for you" to "I'll get it done for you"

The core capability of the previous generation of Apple Intelligence was understanding and generating content. It could summarize long messages, rewrite email tone, and find specific images from the photo library. But it couldn't do things across apps. This is the most fundamental dividing line between the two generations of products.

The new Siri AI leverages the App Intents framework and Spotlight personal data indexing to achieve cross-application contextual understanding and task execution. Apple's official website provides several typical scenarios: Siri can extract hotel booking confirmation numbers from emails, find and directly book restaurants recommended by friends in the Messages app, and even automatically retrieve verification codes from emails and display them on the Phone app interface during a call.

This feature, called Call Context, allows users to directly access the required number from the call screen without switching screens or manually searching. It addresses a specific and frequent pain point: the hassle of simultaneously checking text messages or emails while on a call.

The upgrade to the password management app also points to this point. The new Passwords app can not only detect weak and duplicate passwords, but also automatically navigate users to the corresponding website, complete the login, generate a strong password, and save it. The tediousness of changing passwords is well-known, and most users choose not to. Apple entrusts this task to an AI agent, reducing the user's security maintenance costs.

Safari has also incorporated similar capabilities. The new Notify Me feature can monitor changes to specified web pages, such as a product price drop or a page being restocked, and then proactively push notifications. Users can simply wait in the background without having to manually refresh the page every few days. Another more thorough feature allows users to describe their needs in natural language, enabling AI to directly generate a custom browser plugin. For example, a user could say, "Help me automatically highlight rows in all tables on this webpage that have amounts exceeding 500," and Safari will generate the corresponding plugin to accomplish this. This essentially opens a zero-code channel for ordinary users to customize their browser.

The Shortcuts app has also undergone the same natural language transformation. "Describe a Shortcut" allows users to simply type "Automatically announce today's calendar and weather at 8 AM every morning," and AI can automatically assemble the corresponding shortcut steps. These changes all point to the same goal: to make system functions that were originally only available to advanced users or developers accessible to ordinary users who can do them with just a few words.

With Siri installed on your camera, expand and reconstruct perspective images together in your photo album.

Visual Intelligence has significantly expanded its coverage in this update. A new "Siri Mode" has been added to the Camera app, allowing users to point at restaurant bills and have Siri calculate the cost per person (split the bill). Siri then recognizes the image content and initiates the splitting request directly through Apple Cash. Pointing at a book to check reviews or at food to display nutritional information—these scenarios highly overlap with the functional path of Google Lens over the years.

The difference lies in the fact that Apple has extended this visual understanding capability to more devices. Screenshots taken on an iPad can be analyzed directly by Siri, while on a Mac they can be accessed via keyboard shortcuts, and Vision Pro can also utilize it. The camera is no longer the sole entry point; Apple is building a universal visual understanding layer that covers all devices.

The Photos app also received two new AI-based features: Spatial Reframing and Extend. Extend involves AI automatically generating additional content around the edges of a photo. Spatial Reframing, however, is fundamentally different. It uses spatial computing models to alter the perspective of a photo, not simply cropping or stretching, but recalculating the spatial relationships of objects within the image. Apple demonstrated this at the launch event with a side-view photograph of a building, which, after processing, became a frontal view. While the official processing effect on non-architectural scenes has not yet been released, the technical approach suggests that this involves an additional step of spatial understanding calculation compared to ordinary generative extend.

The upgrade to Image Playground takes another dimension. The first generation could only generate cartoon or illustration-style images, but this time it directly supports photorealistic image generation. Due to the high computational demands, Apple runs it on Private Cloud Compute, rather than on the device. Each generated image is forcibly embedded with a SynthID invisible watermark to identify AI-generated content. This feature also has a daily usage limit; the specific quota will vary depending on the iCloud+ subscription plan in the official version, and the exact number has not yet been disclosed.

Hardware barriers and regional restrictions are unavoidable realities.

The scope of support for AI features is much narrower than that for system updates.

iOS 27 compatibility starts with the iPhone 11, but core Apple Intelligence and Siri AI functionality is limited to iPhone 15 Pro and later models, and iPads and Macs with M1 and later chipsets. Some more advanced features, such as more expressive voice and advanced dictation, require newer hardware: iPhone Air, iPhone 17 Pro series, or M4 iPads and M3 Macs with 12GB or more of the same memory.

Regional restrictions are equally stringent. Due to the EU's Digital Markets Act, Siri AI is temporarily unavailable on iOS and iPadOS in the EU. Apple Intelligence is also temporarily unavailable in mainland China due to regulatory requirements. AI functions based on the PCC cloud, such as photorealistic image generation, are also locked in markets with regional restrictions. For users accustomed to Apple's globally unified experience, this dual approach of separating functions by region and device may be confusing.

Apple's AI journey this year: Delays, lessons learned, and a change in the list of partners.

When Apple first introduced the Apple Intelligence concept at WWDC 2024, Siri was described as an intelligent assistant capable of deeply understanding personal data and performing tasks across apps. However, these features were subsequently delayed several times, leading to a class-action lawsuit filed by users for false advertising, which Apple ultimately settled for $250 million.

This experience directly impacted the density of content and the level of caution in the promises made in this release. The promises made a year ago must now be put on the table.

Throughout 2025, Apple significantly slowed down the release of consumer-grade AI features, a period dubbed Apple's "Gap Year" for AI. During this time, Apple restructured its AI division internally in preparation for a major overhaul in 2026. According to Yahoo Finance, Apple's capital expenditure plan for 2026 is $14 billion. Compared to Amazon and Microsoft's AI infrastructure spending of hundreds of billions of dollars, this figure appears restrained. Apple does not participate in the computing power arms race, relying instead on its own chip ecosystem and edge computing for differentiation.

The most noteworthy change is the shift in partners. In 2024, Siri integrated with OpenAI's ChatGPT external model. By 2026, Apple had established a long-term, deep partnership with Google Gemini, utilizing Gemini technology to custom-develop the next generation of Apple Foundation Models. This shift from a single collaboration to multi-party parallelism, and from direct invocation to deep customization, reflects Apple's unwillingness to be tied to a single third party for its foundation models. For users, this means that the model capabilities behind Siri are more diverse, and it's possible to switch between different models to leverage their strengths for specific tasks.