Apple Intelligence and the new Siri
The most-watched front of the event. This is where Apple's AI credibility is decided: how real the new Siri is, which features actually ship and when, what runs on-device and what in the cloud. I gather the Apple Intelligence announcements here with a critical eye on the gap between promise and actual availability.
In the post-keynote Q&A, Craig Federighi clarified how Siri AI reads on-screen content in real time and builds context across multiple apps, without sending data off the device.
MacRumors published on June 10 first hands-on videos of Siri AI on iOS 27 and macOS Golden Gate, showing how the new assistant actually behaves compared to keynote promises.
MacRumors spotted Apple readying the genai.apple.com subdomain — not yet live, but signaling a separate communications strategy for its generative AI products.
Beyond voice and Dynamic Island, Siri AI introduces a specific gesture to switch to text input: pulling down the response tab opens a text field with deeper information. The World Cup demo showed how it works in practice.
On stage at WWDC, Craig Federighi used unusually direct language to differentiate Apple's approach from competitors: those chasing AI for its own sake, without regard for users, are on the wrong path. A rare competitive positioning move for Apple.
In a post-keynote tech talk at Apple Park, Craig Federighi explained the reasoning behind launching Siri AI as a dedicated chatbot-style app, after years of Apple publicly distancing itself from that format.
Analyst Ming-Chi Kuo commented on WWDC 2026, noting that short-term market reaction is irrelevant: Apple's real challenge is demonstrating it can use Gemini models better than Google itself, through tight hardware and ecosystem integration.
Post-keynote analysis reveals an internal hardware stratification within Siri AI: the most advanced features — customizable expressive voices and dictation accuracy leap — require iPhone Air, iPhone 17 Pro or Pro Max, iPad with M4 or later chip, and Mac M3 or later with at least 12GB of unified memory.
Although iOS 27 ships in fall to compatible devices, Siri AI will not be included at launch: Apple plans a separate beta slipping to a later point in 2026, decoupling the new assistant from the system update.
Safari's new Organize Tabs feature uses Apple Intelligence to analyze open tab content and group them by topic — research, shopping, work — without manual intervention.
On macOS Golden Gate, Siri AI is integrated directly into Spotlight. Conversations can then be resumed on another device — iPhone or Mac — without losing context.
With iOS 27, Siri AI displays rich information cards on screen and can handle multiple simultaneous requests in a single conversational turn.
The second Apple Foundation Model — the next-generation multimodal one — doesn't run on all Apple Intelligence-compatible devices. It requires at least 12GB of RAM and an A19 Pro on iPhone, M4 on iPad, M3 on Mac.
During the WWDC 2026 keynote, the Siri AI demo showed how users can ask for directions to a location seen in an Instagram post, combining Visual Intelligence and contextual understanding. A capability Google has offered for some time, now integrated into Apple's ecosystem.
Vice President Mike Rockwell led the Siri AI demo during the WWDC 2026 keynote, showing among other things how to ask for directions based on an Instagram post. A new face for a product Apple positioned as its most anticipated overhaul in years.
With iOS 27, Siri AI allows adjusting not only the chosen voice but also the speech rate and expressiveness level, adapting the assistant's conversational style to personal preferences or the user's accessibility needs.
On macOS Golden Gate, the screenshot tool gains Visual Intelligence capabilities: users can frame a festival schedule, select events of interest, and add them to the calendar directly via Siri AI, without copying and pasting text.
The dedicated Siri AI app is not just a gateway to the voice assistant: Apple confirmed it supports document and file analysis, image generation, and advanced web search, bringing the experience closer to standalone chatbots like ChatGPT.
All images generated or edited with Apple Intelligence in iOS 27 will automatically receive a SynthID watermark, Google DeepMind's standard for identifying synthetic content. The watermark is invisible to the eye but detectable by analysis tools.
Apple expands Visual Intelligence with two practical features: the Camera can read food nutrition labels and log data to the Health app, or extract phone numbers and addresses from business cards and add them directly to Contacts.
Alongside the relaunch of Image Playground with new styles and realistic images, Apple confirmed that images generated by the app are excluded from any use in training its models.
Developer betas of iOS 27 and iPadOS 27 are available since keynote day, but Siri AI is behind a waitlist: the OS downloads fine, but the headline feature is not yet accessible to everyone.
Apple Intelligence enhances Accessibility Reader with the ability to summarize long texts and translate them, making written content more accessible to people with reading difficulties or non-native speakers.
Apple updated its press release on June 9: Siri AI in Europe will be available initially only on macOS 27 and visionOS 27, excluding watchOS 27 contrary to original statements. A detail that redraws the European map of AI feature access.
In the iOS 27 Mail app, Apple Intelligence automatically identifies actionable information in email bodies — dates, times, addresses — and presents them as quick tap-to-execute actions directly in the message, without leaving the app.
iOS 27 introduces AI suggestion icons inside Messages and Mail: the system recognises actionable information in conversations — dates, places, tasks — and offers actions like adding a calendar event or creating a reminder without leaving the app.
Apple has upgraded the Clean Up tool in the iOS 27 Photos app, making object removal more accurate through updated generative models. The feature joins Extend, Reframe, and Enhance as the fourth pillar of AI editing.
Unlike Siri AI — which launches in English only — Apple Intelligence arrives this fall supporting a broad set of languages including Italian, French, German, Spanish, Japanese, Portuguese, Simplified and Traditional Chinese, Korean, Swedish, Turkish, Vietnamese, and more.
iOS 27 redraws the relationship between Siri and third-party models. Where Siri previously handed off queries to ChatGPT with an explicit transfer, Gemini now powers responses directly inside Siri AI. OpenAI retains a presence in the ecosystem but loses the privileged position built with iOS 26.
Apple rebuilt the search engine in Photos, Spotlight and Mail with a new index that processes content nearly in real time. Unified search is no longer siloed by app: it becomes a common access point to all device content.
Apple integrates Apple Intelligence into accessibility features: VoiceOver gets richer descriptions via Image Explorer, the Action Button becomes a trigger for live questions about what the camera sees, and Magnifier responds to natural language voice commands.
With iOS 27, Apple Intelligence no longer waits for an explicit request: it reads conversations, emails, and the photo library to automatically suggest creating events, reminders, or sharing relevant images. A shift from reactive assistant to contextual agent.
Apple drew a line within its own lineup: the most advanced Siri AI features — expressive voices and enhanced dictation — require iPhone 17 Pro, 17 Pro Max, or iPhone Air. Other compatible devices get a reduced version.
Swiping down from the top center of the screen opens 'Search or Ask', a unified panel from which to launch system searches, run shortcuts or delegate complex queries to ChatGPT, Claude or Gemini. It is the interface that replaces the old Spotlight bar as iOS 27's central hub.
On iPad, Visual Intelligence integrates directly into the screenshot workflow: capture a screen and AI analysis is immediately available. In parallel, Siri AI lands in Spotlight with the same depth of response and personal context awareness already seen on Mac.
Apple introduces a daily cap on image generation in Image Playground. Those wanting to unlock more capacity will need to do so through higher iCloud+ subscriptions — a first concrete signal of how Apple intends to monetize its most resource-intensive AI features.
With 'Describe a Shortcut', users explain in plain language what they want to automate and Apple Intelligence builds the steps automatically — lowering the barrier to one of the platform's most powerful tools.
In a Q&A with press immediately after the keynote, Craig Federighi clarified the philosophy behind Siri AI: not an alternative chat window, but an assistant deeply rooted in the system experience.
With iOS 27 and macOS Golden Gate, Apple Intelligence introduces automatic grammar and style checking active in nearly every application, without manual activation. It is not autocorrect: it is a deeper layer of analysis.
With iOS 27 and macOS Golden Gate, Image Playground expands beyond standalone images: it can now create custom wallpapers and Contact Posters, and new APIs let developers integrate the same capabilities into their own apps.
With Spatial Reframing, the Photos app in iOS 27 and macOS Golden Gate analyzes the spatial depth of an image and allows its perspective to be changed, not just its crop. A feature distinct from Extend and Clean Up, closer to the domain of computational photography.
macOS Golden Gate upgrades Writing Tools with deeper contextual capabilities: the system learns the tone and style a user adopts with each contact and adapts generated drafts accordingly. Writing feedback is also new.
macOS Golden Gate brings Visual Intelligence out of the Camera app and makes it accessible anywhere on the desktop via a keyboard shortcut, similar to the screenshot tool. Results can continue directly in the new Siri app.
At the launch of iOS 27 and iPadOS 27, Siri AI will not be available in either the European Union — due to the Digital Markets Act — or in China. On macOS Golden Gate, watchOS 27 and visionOS 27, the European block does not apply.
Apple has unveiled a second version of its Apple Foundation Models, this time multimodal: the model can understand speech, read text, and interpret images, all on the user's hardware without cloud involvement.
Apple confirmed during the keynote that Siri AI will initially be available in English only. Support for additional languages will follow, with no specific date announced.
With Call Context in iOS 27, Apple Intelligence surfaces contextually relevant information during a phone call — contact details, related emails or messages — without the user having to search manually. A silent assistant active while you talk.
With "App Actions" in iOS 27, Apple Intelligence can execute operations across multiple apps in sequence — suggest an event from a message, surface photos from a conversation, trigger contextual actions — without the user manually initiating anything. This is the automation Apple had been promising since WWDC 2024.
iOS 27 takes Writing Tools well beyond basic rewriting: a dedicated AI grammar checker arrives, Smart Reply suggestions in the user's own writing style, and a "Write with Siri" toggle above the keyboard. The goal is to reduce time spent formulating text.
In iOS 27, Genmoji produces higher-quality images thanks to a new model and introduces "Suggested Genmoji": personalised emoji automatically suggested based on the user's message and media history.
With the new Siri AI app, all chat history syncs across iPhone, iPad, Mac, Apple Watch, and Vision Pro, allowing users to continue a conversation on whichever device is most convenient without losing context.
With iOS 27, Apple moves Visual Intelligence from the Camera Control button (exclusive to iPhone 16) into the Camera app itself as a dedicated mode, making it accessible to all Apple Intelligence-compatible devices.
In iOS 27 and macOS Golden Gate, the Home app uses Apple Intelligence to summarize security camera footage in text and to group frequent notifications into ongoing activities instead of individual alerts.
Apple announced a second tier of its Apple Foundation Models, called AFM Cloud Pro, designed for the most demanding tasks and comparable in quality to Gemini Frontier models. It runs on Nvidia GPUs in Google's cloud infrastructure.
With Siri AI, Apple introduces more natural-sounding voices and new customization parameters: users can adjust voice pace and degree of conversational expressivity.
In Safari on iOS 27 and macOS Golden Gate, Apple Intelligence automatically groups tabs by topic and can alert users when a specific web page changes — such as a price drop or product restock.
The new Siri AI extends beyond iPhone and iPad to Apple Watch with a new app grid, CarPlay, and AirPods, with a customizable voice pace and expressivity.
Apple has deeply updated Image Playground: image generation moves away from the cartoon aesthetic and embraces far more realistic styles. New formats and sizes arrive too, along with the ability to apply effects to existing photos.
iOS 27 brings Apple Intelligence into everyday apps: Messages suggests calendar events, Mail surfaces contextual actions, Phone displays relevant info during calls, Home reduces notification noise, and Passwords upgrades weak credentials.
Apple confirmed collaboration with Google to develop the new Foundation Models powering Apple Intelligence. The new Siri will not be available in the EU on iOS and iPadOS at launch, citing the Digital Markets Act.
iOS 27 brings three new AI-assisted editing functions to the Photos app: Extend expands image borders, Enhance optimizes quality, and Reframe changes perspective in spatial photos.
With iOS 27, Siri gets its own standalone app, a redesigned conversational interface, and the ability to set third-party AI assistants as the default for Apple Intelligence features.
The new iOS, iPadOS and macOS releases
Xcode, Swift and the tools for those who build
The visual language and interfaces
visionOS and spatial computing
Health, fitness and watchOS
iCloud, Continuity and the ecosystem
Privacy and security