Apple Intelligence is Apple's personal AI system built into iPhone, iPad, and Mac — a suite of AI features focused on writing assistance, image generation, smarter Siri, and on-device privacy that processes most requests locally rather than sending them to the cloud.
Apple was conspicuously quiet during the first wave of the AI moment. While OpenAI launched ChatGPT, Google raced to ship Gemini, and Microsoft embedded Copilot into everything, Apple said almost nothing. Then at WWDC 2024, they unveiled Apple Intelligence — and the approach made clear why they had waited. Apple wasn't trying to ship first. They were trying to ship differently.
The privacy angle is genuine, not just marketing. Most of what Apple Intelligence does happens on your device, using models that run locally on Apple Silicon chips. What can't run locally goes through Private Cloud Compute — a system Apple has described in unusual technical detail and opened to external security auditors. For users who care about where their data goes, this matters in ways that competing AI systems don't address as directly.
Here's what Apple Intelligence actually does, what it can't do, and how it fits into the broader AI landscape.
1. What Is Apple Intelligence?
Apple Intelligence is the name Apple uses for the collection of AI features integrated into iOS 18, iPadOS 18, and macOS Sequoia, announced at WWDC 2024 and rolled out progressively through 2024 and 2025. It's built on Apple's own on-device language models, optimized for Apple Silicon, with cloud processing handled through Private Cloud Compute when local processing isn't sufficient.
The features span several categories: writing tools that work system-wide across any app, image generation, a significantly upgraded Siri, notification summarization, and integration with ChatGPT for queries that benefit from a more capable external model. More information is available at apple.com/apple-intelligence.
Apple Intelligence requires an iPhone 15 Pro or iPhone 16 or later, iPad with M1 chip or later, or Mac with M1 chip or later. The hardware requirement exists because the on-device models need the Neural Engine in Apple Silicon to run efficiently. Older devices, even capable ones, don't have the chip architecture required.
2. The Privacy Architecture
The most technically distinctive aspect of Apple Intelligence is how it handles the tension between AI capability and privacy — worth understanding in detail because it represents a genuinely different architectural approach from every other major AI system.
Most AI processing in Apple Intelligence happens entirely on your device. The language models run locally on the Neural Engine in Apple Silicon chips, meaning your prompts and data never leave your phone or computer for routine tasks. Writing suggestions, notification summaries, photo editing — all processed locally with no network request.
For more complex requests that exceed what on-device models can handle, Apple routes them through Private Cloud Compute — Apple's cloud AI infrastructure. PCC is architecturally unusual: Apple has committed that even Apple employees cannot access the data processed there, that the servers run verifiable software that security researchers can inspect, and that requests aren't logged in ways that could be tied to individuals. Apple opened the PCC technical specifications and a virtual research environment to external security researchers — a level of transparency unusual for any cloud AI system.
For queries that benefit from a more powerful model — where neither local processing nor PCC is sufficient — Apple integrates with ChatGPT through an opt-in connection that asks for permission before sending any request to OpenAI, and doesn't send requests without explicit user consent.
3. Key Features of Apple Intelligence
Writing Tools
Available system-wide — in Mail, Messages, Notes, Pages, and any third-party app that uses the standard text input. Proofread catches grammar and spelling errors with context-aware suggestions. Rewrite changes the tone or style of selected text. Summarize condenses long text into key points. These tools are accessible via the right-click context menu on any selected text, making them genuinely integrated rather than requiring a separate app.
Priority Notifications and Summaries
Apple Intelligence summarizes notification stacks — a group of messages from a thread appears as a single summary rather than ten individual alerts. It also surfaces Priority Notifications that it determines are time-sensitive or important, surfacing them at the top of the notification list. In Mail, an intelligent priority inbox highlights messages that likely need a response.
Image Playground
A dedicated app for generating images in three styles — Animation, Illustration, and Sketch — from text descriptions or photo suggestions. Not photorealistic, and deliberately so — Apple positioned Image Playground for fun, expressive images rather than trying to compete with Midjourney or DALL-E on realism. Images generate quickly on-device.
Genmoji
Generate custom emoji from text descriptions. Type "a golden retriever wearing sunglasses at the beach" and Genmoji creates a custom emoji in Apple's emoji style. Available in Messages, they function like standard emoji and can be sent to anyone.
Image Wand (iPad)
Draw a rough circle with Apple Pencil in the Notes app and Apple Intelligence generates an image based on surrounding context. Useful for illustrated notes and quick visual references without leaving the app.
Photos Enhancements
The Photos app gained several AI features: Clean Up removes unwanted objects from photos (similar to Google's Magic Eraser), Natural Language Search lets you find photos by describing them in plain language ("photos from the beach last summer"), and Memory Movies generates video compilations from your photo library based on a theme you describe.
Upgraded Siri
Siri received its most significant overhaul since its 2011 launch. It now understands context across a back-and-forth conversation, can take actions within apps (not just launch them), has access to your personal context including contacts, calendar, and messages, and — when connected to ChatGPT — can handle more complex knowledge queries. The new Siri also understands what's on your screen, allowing requests like "send this article to my sister" when viewing content.
4. Apple Intelligence and ChatGPT
The integration with ChatGPT is one of the more nuanced aspects of Apple Intelligence. Apple built its own models for on-device processing, but rather than trying to build a cloud AI model that competes with GPT-4o, it partnered with OpenAI to handle queries that benefit from a more capable external system.
The opt-in design is deliberate. When a query would benefit from ChatGPT, Siri asks whether you want to send it before doing so — you can decline and get a less capable response, or approve and have it sent to OpenAI's servers. Apple explicitly doesn't send requests without this permission, and the integration doesn't require an OpenAI account or subscription (though Plus subscribers can connect their account for access to more advanced features).
This architecture reflects Apple's positioning: they're not trying to be in the AI model business. They're trying to be in the AI experience business — designing the interface, managing the privacy architecture, and integrating the best external models when needed, rather than competing to build the most capable model themselves.
5. What Apple Intelligence Cannot Do
The on-device focus creates real limitations worth being clear about.
Apple Intelligence's writing and language capabilities are less capable than ChatGPT, Claude, or Gemini on complex tasks. The on-device models are specifically optimized for the features they power — summarization, rewriting, tone adjustment — rather than for open-ended conversation or complex reasoning. For extended AI conversations, you're using ChatGPT through the integration, not Apple's own models.
There's no equivalent to Claude's extended thinking, ChatGPT's deep research mode, or Gemini's very long context window. Apple Intelligence is designed for ambient, integrated AI assistance — things that happen quickly in the background — not for sitting down and having a long AI-assisted work session.
And the hardware requirement means a significant portion of iPhone users can't access it at all. iPhone 15 standard, iPhone 14, and older devices are excluded. This is a meaningful limitation given that many people keep phones for three to four years.
6. Apple Intelligence vs Google's AI Features vs Samsung Galaxy AI
| Apple Intelligence | Google AI (Pixel) | Samsung Galaxy AI | |
|---|---|---|---|
| On-device processing | ✅ Primary approach | ⚡ Mixed | ⚡ Mixed |
| Privacy architecture | ✅ PCC, auditable | ⚡ Standard cloud | ⚡ Standard cloud |
| Writing tools | ✅ System-wide | ✅ System-wide | ✅ Good |
| Image generation | ✅ On-device | ✅ Cloud-based | ✅ Cloud-based |
| External AI integration | ✅ ChatGPT opt-in | ✅ Gemini native | ⚡ Limited |
| Cost | ✅ Free with device | ✅ Free with device | ✅ Free (some features) |
The honest comparison: Google's AI features on Pixel phones are more capable in some areas because Google isn't constrained by the same on-device-first philosophy. Apple's approach is more private and arguably more coherent as a user experience. Samsung's Galaxy AI is the most broadly deployed across different price points but relies heavily on cloud processing. Which matters most depends on whether you prioritize capability, privacy, or accessibility.
7. How to Use Apple Intelligence
If you have a compatible device running iOS 18 or later, Apple Intelligence is available in Settings → Apple Intelligence & Siri. You may need to join a waitlist depending on your region and device — Apple has been rolling it out progressively.
The features don't require any separate download or account. Writing Tools appear automatically when you select text in any app — look for the floating toolbar. Notification summaries appear by default on the lock screen. Image Playground is a standalone app. Siri improvements are active once Apple Intelligence is enabled.
The ChatGPT integration is opt-in: go to Settings → Apple Intelligence & Siri → ChatGPT and enable it. You can connect an existing OpenAI account for Plus-tier access or use the basic integration without an account.
Conclusion
Apple Intelligence is a different kind of AI product than ChatGPT or Gemini — less about having a conversation with AI and more about AI quietly making your existing apps and device more useful. The privacy architecture is the most technically serious attempt by a major consumer technology company to address the data questions that most AI systems sidestep.
Its limitations are real: the on-device models aren't as capable as cloud frontier models for complex tasks, the hardware requirement excludes a lot of users, and the rollout has been slower than Apple's ambitions. But the direction — ambient, private, integrated AI that works within your existing workflow rather than asking you to switch to a new one — represents a coherent and distinctive vision for what consumer AI should be.
FAQ
Q: Which iPhone models support Apple Intelligence?
A: Apple Intelligence requires iPhone 15 Pro, iPhone 15 Pro Max, or any iPhone 16 model (including the standard iPhone 16). Earlier iPhones including iPhone 15 standard and iPhone 14 series are not supported due to the Neural Engine requirements of the on-device AI models.
Q: Is Apple Intelligence free?
A: Yes, Apple Intelligence is included at no additional cost with supported devices running iOS 18, iPadOS 18, or macOS Sequoia. The ChatGPT integration is available without an OpenAI account for basic use; connecting an existing ChatGPT Plus account unlocks more advanced features.
Q: Does Apple Intelligence send my data to Apple's servers?
A: Most Apple Intelligence processing happens on-device and never leaves your device. More complex requests are processed through Private Cloud Compute — Apple's cloud AI infrastructure designed so that even Apple cannot access the data. Requests to ChatGPT are opt-in and require explicit permission before any data is sent to OpenAI.
