What the New Siri Is: Apple Intelligence, Not a Gemini Clone
Apple’s new Siri is an iOS AI assistant built on Apple’s own Apple Foundation Models and system-level intelligence, combining on-device AI processing, private cloud services, and personal context to deliver conversational help that feels integrated into the operating system instead of behaving like a standalone chatbot. That description matters because many observers assumed that Apple’s WWDC 2026 announcements meant Siri AI capabilities would mirror Google’s Gemini experience. Craig Federighi pushed back on that view in post-keynote briefings, stressing that the iOS assistant is not a thin wrapper over Google’s technology. He explained that Apple does not ship the Gemini app, does not run Google’s client code, and does not rely on Google Assistant or Google Search to answer questions. Instead, Siri sits on top of Apple Intelligence, a tightly integrated stack that is being treated as a core system feature rather than a bolt-on service.
How Apple Gemini Integration Shapes, But Does Not Define, Siri
The new Siri does use Google Gemini, but in a limited and behind-the-scenes way. Apple’s AI leadership says that Gemini frontier models helped refine Apple’s own Apple Foundation Models through reinforcement learning, giving Siri a training partner without importing Google’s deployment stack. According to PCMag’s WWDC coverage, Amar Subramanya described a family of five AFMs, from AFM Core on-device to AFM Cloud Pro for complex, agent-like tasks. A system orchestrator decides which model to use, based on the complexity of each request. When a request requires more power than a device can provide, Siri sends it to Private Cloud Compute, Apple’s privacy-focused server environment. Here, Apple even relies on Nvidia GPUs in Google’s cloud, but still keeps user data within its own controlled infrastructure. In short, Apple Gemini integration helps teach Siri, while Apple’s own stack runs Siri in production.
On-Device AI Processing, Privacy, and Personal Context
Apple is betting on on-device AI processing as the core of Siri’s new identity. AFM Core and AFM Core Advanced sit directly on Apple silicon, handling lightweight queries, expressive voice commands, dictation, and native multimodal input without leaving the iPhone. This approach allows Siri to work with personal context such as messages, documents, and app data inside a “System Orchestrator” framework that keeps processing tightly linked to iOS. When cloud help is needed, Private Cloud Compute extends that model, designed so that, as Federighi described, not even Apple can read user requests. That combination of local intelligence and locked-down cloud routes enables more contextual and personalized responses while still upholding Apple’s long-standing privacy stance. For users, the net effect is an iOS AI assistant that feels informed by their digital life but less exposed to outside data collection than typical chatbot-style services.
New Siri App, Visual Intelligence, and a Rebuilt Assistant Experience
Beyond the backend, Apple is overhauling the way people work with Siri. The company is rolling out a dedicated Siri app that turns the assistant into a destination, not just a voice layer over the Home button or Dynamic Island. Siri can now handle richer, multi-step requests using AFM Cloud Pro, acting as an agent that understands earlier context in a conversation and coordinates actions across apps. Visual Intelligence features add multimodal understanding, so Siri can interpret on-screen content or images as part of a request and respond with more precise help. Together, these changes move Siri AI capabilities well beyond the handoff model users saw when it prompted for ChatGPT responses. Apple wants Siri to be the front door into Apple Intelligence itself, fusing language, vision, and app control into a single, persistent, system-level assistant.
