Siri: Google AI Update
Apple integrates Google's AI model Gemini into the next Siri version and pays for it about 1 billion USD per year. This complements the existing ChatGPT integration and Apple's own Apple Intelligence. The decision to bring Google on board after OpenAI raises questions about the impact on users, data, and everyday life with Siri.
Source: Own Representation
Apple's AI Strategy
Siri, Apple's voice assistant, was introduced in 2011 with the iPhone 4s and is today available on many Apple devices. Originally designed for simple commands, Siri often appeared stiff and not very ‘conversational’ ( theguardian.com).
Apple Intelligence is Apple's own AI layer, which offers features such as text summarizations, email sorting, and image editing. Many of these features run directly on the device, more complex computations are outsourced to Apple's “Private Cloud Compute” (PCC) ( apple.com). PCC is designed so that requests are encrypted, processed briefly, and then discarded, without employees or partners having access to the contents ( security.apple.com).
ChatGPT from OpenAI is a well-known AI chatbot. Through Apple Intelligence, Siri can access ChatGPT when needed, for example for creative ideas or complex explanations ( apple.com). By default, Siri asks for permission before sending a request to OpenAI, and IP addresses are obfuscated. The integration currently uses GPT-4o and is planned to be updated to GPT-5 with iOS 26 ( theverge.com).
Google Gemini is Google's family of large language models that can understand and generate text, code, images, audio and video. Gemini is integrated into many Google products and can function as a background assistant that plans, explains, and brings information together ( gemini.google).
Reports suggest that Apple licenses a specially customized, extremely large Gemini variant (1.2 trillion parameters) from Google to power the next generation of Siri ( reuters.com). This model is intended to run on Apple's own Private Cloud Compute infrastructure, not directly in Google Cloud ( macrumors.com).
In June 2024, Apple unveiled Apple Intelligence at WWDC, a bundle of new AI features with an upgraded Siri ( apple.com). Shortly thereafter, the partnership with OpenAI was announced, enabling Siri to access ChatGPT ( theguardian.com).
In early 2025, Apple pushed the promised 'AI-Siri' features to 2026, as the first version 'was not reliable enough' ( theverge.com).
Meanwhile, there were reports of Apple's talks with Google and other AI providers. In autumn 2025, the information intensified that Apple would choose Google and use a specially customized Gemini model with 1.2 trillion parameters, for which about 1 billion USD per year would be paid to Google ( reuters.com).
The Gemini model is intended mainly for summaries and multi-step task planning ( macrumors.com). Apple's own models remain responsible for many Siri functions, especially on-device. Gemini runs on Apple's Private Cloud Compute ( security.apple.com).
The reports point to a major Siri update around iOS 26.4 in spring 2026, with initial rollouts around March 2026 on the iPhone 16 and newer models ( macrumors.com). At the same time, the ChatGPT integration is expected to switch from GPT-4o to GPT-5 ( theverge.com).
Source: arstechnica.com
The introduction of Apple Intelligence at WWDC 2024 marks a turning point in Apple's AI strategy.
Backgrounds & Motives
Apple's decision to bring Google on board after OpenAI appears paradoxical at first glance. The reasons are varied:
Speed and pressure to perform: Training large language models is expensive and compute-intensive. Apple has developed its own models, but these lag behind the performance of the Gemini variant intended for Siri ( macrumors.com). To quickly catch up with the best assistants, using a strong third-party model is understandable.
Role of Apple Intelligence: Apple pursues a hybrid strategy: as much as possible on the device, the rest in the cloud under strict privacy rules ( security.apple.com). ChatGPT acts as a 'super-expert' outside of Apple's infrastructure ( help.openai.com). For Siri's core logic, which understands context and plans actions, a licensed Gemini core running on Apple servers fits the architecture better.
Power balance between AI providers: Exclusive reliance on ChatGPT would make Apple dependent on a single company. Cooperation with Google and the further development of its own models create negotiating room and technical redundancy ( macrumors.com).
Business relationship with Google: The Gemini deal fits into the long-standing business relationship between Apple and Google, where Google already pays billions to be the default search engine on Apple devices ( macrumors.com). Google gains additional reach and revenue, Apple premium devices, without having to build datacenters to the same extent. Bloomberg and Reuters describe the Gemini model as a 'gap-filler' until Apple's own models are strong enough ( reuters.com).
User expectations: After ChatGPT and Gemini, users expect an assistant to hold longer conversations, plan several steps, and understand content ( techradar.com). The partnership with Google is intended to meet these expectations.
Source: YouTube
Source: maginative.com
The partnership between Apple and Google (Gemini) is a central pillar of the new Siri era.
Impacts on users
For users of a current iPhone with Apple Intelligence (e.g., iPhone 15 Pro, iPhone 16 and newer), the upcoming Siri update means that behind the same voice there could be several 'brains' at work: a local Apple model, an Apple Cloud model, a Gemini model, and, if needed, ChatGPT ( apple.com). This should enable Siri to conduct longer, context-aware conversations, plan complex task chains in and between apps, and summarize content from various sources ( macrumors.com).
Three key points for daily life:
1. The settings: Under 'Apple Intelligence & Siri', you can already control today whether ChatGPT is used and what confirmations are required ( support.apple.com). It's likely that Apple will offer similar transparent switches for Gemini.
2. Source checking: Generative AI can be mistaken. For important facts (health, finance, law), it is advisable to cross-check the answers with reputable sources, such as official Apple pages, authorities, or specialist portals ( apple.com).
3. Your personal comfort level: Apple's approach with on-device models, Private Cloud Compute, and optional third-party models enables conscious decisions. Users can decide whether they want maximum intelligent automation or limit AI to the essentials and not route sensitive contexts through assistants ( lifewire.com).
Source: mashable.com
The improved Siri will make everyday tasks more intuitive and efficient.
Open Questions & Future
Despite the many reports, some core questions remain open. It is unclear how transparent Apple will be in the interface about which model is answering. With ChatGPT, it is currently visibly indicated when an answer comes from OpenAI; whether there will be a similar indicator for Gemini is not yet documented ( support.apple.com).
The long-term strategy is also not fully visible. Reports indicate that Apple is working on its own cloud model with around 1 trillion parameters, which could be ready for use from 2026 ( macrumors.com). Whether Gemini will then disappear entirely, remain only in certain markets or scenarios, or be complemented by further third-party models, is left open by Apple.
From a privacy-law perspective, additional questions arise: How exactly will models developed by a third party but running on Apple infrastructure be audited? What guarantees do regulators have that no hidden data flows occur? Apple's commitments to Private Cloud Compute and OHTTP techniques are ambitious, but must prove themselves in independent audits ( security.apple.com).
Finally, the question remains how the balance of power among the major AI providers will evolve. OpenAI, Google and Apple pursue their own business models, and regulators monitor exclusive deals closely ( nytimes.com). It is possible that courts and competition authorities will influence the admissibility of AI integrations.
Source: YouTube
In summary, Siri will be less a single assistant and more an interface for multiple AI engines: a small Apple model on the device, a larger Apple model in Private Cloud Compute, a licensed Gemini core from Google, and an optionally enabled ChatGPT. The Siri experience is likely to be more powerful and flexible, but also more complex. It is worth carefully checking the Apple Intelligence and ChatGPT settings, trying new features critically, and cross-checking important information. Apple aims to connect top models from partners with strong privacy; success depends on the technology, user trust, and control over data. apple.com).