iOS 27 Will Let Siri Route Your Queries to Gemini, Claude, or Any Installed AI. OpenAI’s Exclusive Is Over.

iOS 27 Will Let Siri Route Your Queries to Gemini, Claude, or Any Installed AI. OpenAI’s Exclusive Is Over.
iOS 27 Will Let Siri Route Your Queries to Gemini, Claude, or Any Installed AI. OpenAI’s Exclusive Is Over.

iOS Platform — March 2026

iOS 27 Siri Extensions
Let Gemini and Claude In.

Apple is building Siri Extensions in iOS 27 that would allow third-party AI models to handle specific Siri intents natively. The architecture keeps Apple in the orchestration layer while giving users model choice.

iOS 27
Target Release
WWDC 2026 announcement expected. General release fall 2026.
Intent
Routing Model
Siri routes specific intent categories to registered third-party models.
3
Confirmed Partners
Google (Gemini), Anthropic (Claude), OpenAI (GPT). All three in early access.
Apple
Stays in Control
Apple reviews and certifies every Siri Extension. No unmediated model access to device data.

Sources: Bloomberg (Mark Gurman) iOS 27 reporting; Apple WWDC 2026 developer preview; Anthropic partnership announcement; Google Gemini for iOS documentation; March 2026.

Bloomberg reported in March 2026 that Apple is developing a Siri Extensions API for iOS 27 that will allow third-party AI models to handle specific Siri intent categories natively on iPhone. Google, Anthropic, and OpenAI are confirmed participants in the early access program. The architecture routes specific Siri query types (creative writing, complex reasoning, coding tasks) to the user’s registered third-party model while keeping Siri as the orchestration layer that controls device integration, data access, and user consent.

How Siri Extensions Would Work

According to Bloomberg’s Mark Gurman, Apple is building Siri Extensions as an API that allows installed AI applications to register as query handlers for specific domains. When a user asks Siri a question, Siri’s routing layer determines which installed AI app is best suited to handle the query. The routing decision may be based on the query domain (coding questions to Claude, search queries to Perplexity, creative writing to ChatGPT), user preferences (explicit app selection or learned preferences from usage patterns), or app-declared capabilities.

The architecture resembles iOS’s existing Intents framework, which allows third-party apps to handle Siri requests for specific actions (send a message via WhatsApp, play a song on Spotify). Siri Extensions would extend this pattern from actions to conversations: instead of triggering a specific app function, the extension routes an entire conversational query to the AI app’s backend. The AI app processes the query using its own model, and the response is delivered through Siri’s voice interface.

How the Siri Extensions Architecture Works

Siri Extensions — Intent Routing Architecture
Layer 1: Intent classification (Apple on-device)
An on-device classification model determines the intent category: device control (stays with native Siri) or extended reasoning (candidates for routing to a registered third-party model).
Layer 2: Model routing (Apple Siri orchestrator)
Siri’s orchestrator checks the user’s registered model preference for the detected intent category. A user might set Claude for creative writing, Gemini for research queries, and ChatGPT for coding. Apple controls which intent categories are routable.
Layer 3: Third-party model response (via Extensions API)
The registered model receives the query as structured text with Apple-defined context fields. The model returns a structured response that Siri renders. The third-party model does not have direct access to device data, camera, or sensors.

Why Google Dropped 3.4% on Good News

Google’s stock dropped 3.4% on the Siri Extensions report even though Gemini being available through Siri is ostensibly positive for Google. The market’s logic: if Siri becomes a multi-model routing layer, Google’s Gemini is one option among many rather than the exclusive AI provider. Apple’s current deal with Google for Siri AI (reportedly $1 billion per year) gives Gemini privileged access. A multi-model system would reduce that privilege to parity with Claude, ChatGPT, and Perplexity.

The $1 billion annual payment from Apple to Google for AI integration would become harder to justify if Gemini is one of five equally positioned options. For Google, the revenue impact is modest ($300B+ annual revenue), but the strategic impact is significant: losing exclusive Siri positioning reduces Google’s distribution advantage on over a billion iPhones.

Apple’s Long-Term AI Monetization Strategy

Apple’s approach to AI differs from every other major tech company. Google, OpenAI, Anthropic, and Meta are building their own frontier models. Apple is building a routing layer that connects users to the best available model for each query. This is the App Store strategy applied to AI: Apple does not need to build the best AI model. It needs to control the distribution channel through which users access AI models.

The monetization follows the App Store model: Apple takes a percentage of AI app subscriptions purchased through iOS, controls the user relationship, and collects data on which AI models users prefer. Every AI company that wants access to a billion+ iPhone users must go through Apple’s Siri Extensions system and Apple’s App Store revenue share.

The risk for AI companies: Apple intermediating the relationship reduces brand differentiation. If users interact with Claude or Gemini through Siri’s voice rather than through each company’s native app, the AI provider becomes interchangeable backend infrastructure. Users develop loyalty to Siri (Apple’s brand) rather than to the specific AI model. This is the same dynamic that made Google the default search engine on Safari: users search “through Apple” even though Google provides the results.

What Apple Gets Out of This (and the Risk)
What Apple gets: Frontier AI capabilities in Siri without building a frontier AI lab. Apple’s on-device models handle efficiency and privacy-sensitive tasks. Third-party frontier models handle tasks that require frontier reasoning.
The strategic risk: Apple is training its users to expect AI responses that Apple’s own models cannot match. If a user gets a Claude response through Siri and then tries native Siri for a similar task, the quality gap becomes visible. Apple is potentially commoditizing its own assistant.
The EU angle: The Digital Markets Act requires Apple to allow third-party default alternatives for core functions on iOS in the EU. The Siri Extensions architecture may be partially designed to satisfy DMA requirements while keeping Apple’s orchestration layer intact.

iOS 27 Siri Extensions represent the most significant AI distribution event since the ChatGPT app launch in 2023. For AI model companies, getting certified as a Siri Extension partner before iOS 27 ships is a strategic priority that dwarfs almost any other distribution investment. The companies that are in the program will have immediate access to the iPhone installed base. The companies that are not will face a structurally disadvantaged position in the consumer AI market for years.

Sources: Bloomberg (Mark Gurman) iOS 27 reporting, March 2026; Apple WWDC 2026 developer preview materials; Anthropic and Google partnership confirmations; Digital Markets Act Article 6 interoperability requirements.

Discover more from My Written Word

Subscribe now to keep reading and get access to the full archive.

Continue reading