Apple shipped iOS 27 on September 14, ending an 18-month saga that saw the rebuilt Siri slip past three separate target dates. The headline feature is a conversational assistant with memory of your messages and photos. The two details that actually matter for anyone running marketing or content through Apple’s ecosystem are buried well past the press release: Siri’s new reasoning is reportedly powered by Google’s Gemini, and separate code in iOS 27 shows Apple built Siri to be swappable for Claude or ChatGPT.
Why this took 18 months longer than promised
Apple first unveiled the AI-powered Siri overhaul alongside Apple Intelligence in 2024. It was supposed to ship as part of iOS 26.4 in March 2026. According to Bloomberg’s Mark Gurman, Apple hit problems during testing that forced a staggered rollout instead, some features pushed to a May update, others held back entirely until iOS 27 in September. That’s not a minor slip. It’s a year and a half between announcement and delivery for a feature Apple had already put its name behind publicly, and it happened because the software wasn’t ready, not because of a strategy change.
Why it matters: the delay is the reason Siri’s new brain isn’t Apple’s own model for most users. Building a competitive large language model from scratch on Apple’s original timeline didn’t work, so Apple went and got one instead.
Siri now runs on Google Gemini
Reporting on the final iOS 27 rollout indicates the new Siri’s reasoning layer is powered by Google Gemini, licensed in to cover the gap Apple’s own models couldn’t close in time. This is the same company whose search engine Apple already pays billions of dollars a year to remain the default on iPhone. Apple is now paying Google again, on a second front, to run the AI feature at the center of its own operating system.
Separately, code found inside iOS 27 shows Siri was also built to be replaceable by Claude or ChatGPT, meaning Apple left itself a door to let users pick a different model entirely, the way you’d choose a default browser or maps app today. Put the two facts together and the picture is consistent: Apple no longer sees “the AI in Siri” as something it must build and own itself. It sees it as a swappable layer, currently Gemini by default, openly built to support alternatives.
Why it matters: if the model answering questions on 1.5 billion active Apple devices can change depending on a licensing deal or a settings toggle, then “being findable to the AI on someone’s iPhone” stops being a single target. It becomes a moving one, and the company whose model is currently in that seat has enormous influence over how your brand gets described to a customer who just asked their phone about you.
What’s actually new in the Siri experience
iOS 27 runs on iPhone 11 and later, including the second-generation iPhone SE, so no devices were cut. The redesigned Siri sits inside a “Liquid Glass” interface that takes over the Dynamic Island and can hold a real back-and-forth conversation, including follow-up questions, instead of resetting after every request. It has contextual awareness of your messages, contacts, and photos, Apple’s own example is Siri retrieving vacation photos because it remembers the trip-planning messages you sent. There’s also a dedicated Siri app for reviewing past conversations, closer to how a chatbot app works than how Siri used to behave.
One caveat: Apple’s most advanced on-device model is reserved for iPhone Air and iPhone 17 Pro and later. Most of the existing install base gets the new interface and the Gemini-backed reasoning, but not the full on-device horsepower Apple is reserving for its newest hardware.
Why Apple went to a rival instead of shipping late
Apple has form here, and it’s worth remembering before assuming this Gemini deal is a one-off. Apple Maps launched in 2012 to widespread ridicule, missing landmarks, broken directions, towns that didn’t exist, and Apple spent the better part of a decade quietly rebuilding it while most users just switched back to Google Maps in the meantime. The company’s instinct when a first-party product underperforms has historically been to keep investing in-house rather than license a competitor’s technology into the core experience. Siri breaks that pattern.
The difference this time is competitive pressure that didn’t exist during the Maps years. In 2012, there was no widely-used alternative interface threatening to make Apple’s product irrelevant while it caught up quietly. In 2026, ChatGPT, Gemini, and Claude are already how tens of millions of people ask questions daily, on their phones, often instead of opening a search engine. Apple couldn’t afford an multi-year quiet rebuild this time because the alternative wasn’t “people tolerate a worse product,” it was “people route around Siri entirely and the assistant layer of the iPhone becomes irrelevant.” Licensing Gemini was the faster fix, even at the cost of running a competitor’s model at the center of Apple’s own operating system.
Why it matters: a company as protective of its own stack as Apple choosing to outsource the reasoning layer of its flagship AI feature is a strong signal about how far behind Apple’s own models actually were, not just a licensing footnote. It also means the pace of improvement in Siri going forward is now partly tied to Google’s roadmap for Gemini, not just Apple’s own release cycle.
How the assistant field actually lines up right now
| Assistant | Owner | Model behind it | Where it lives |
|---|---|---|---|
| Siri (iOS 27) | Apple | Google Gemini, reportedly, with Claude/ChatGPT support built into the code | iPhone, iPad, Mac |
| Google Assistant / Gemini for Home | Gemini, Google’s own model | Android, Google Home devices | |
| Alexa+ | Amazon | Multiple models via Amazon’s Bedrock, including Anthropic’s Claude | Echo devices, Fire TV |
| ChatGPT voice | OpenAI | OpenAI’s own models | Standalone app, integrated into some third-party products |
The pattern across the entire category: almost nobody is running only their own model anymore. Amazon already blends multiple providers inside Alexa+. Apple is doing the same thing with Siri. The assistant your customer is talking to is increasingly a blend of infrastructure from two or three companies wearing one company’s interface, which is exactly why treating “Apple’s AI” or “Amazon’s AI” as a single, stable target for visibility work is already out of date.
What this means if you run marketing through Apple’s ecosystem
For marketers: this is the same shift search marketing went through with AI Overviews, one layer earlier and with less visibility into what’s happening. If Siri’s answers about your brand or product are shaped by Gemini’s training data and retrieval sources today, and potentially by Claude’s or ChatGPT’s tomorrow if a user switches, then being accurately and consistently described across all of those sources matters, not just optimizing for one. Three concrete steps worth taking this quarter:
- Audit how your brand shows up today. Ask Gemini, ChatGPT, and Claude directly what they know about your product, pricing, and differentiators. Note where the answers are wrong, outdated, or missing entirely.
- Fix your structured, citable content first. FAQs, comparison pages, and pricing pages written in clear, directly answerable language get pulled into AI answers more reliably than marketing copy built around a slogan.
- Treat GEO as infrastructure, not a one-off project. Because the model behind Siri can change with a licensing deal, this isn’t a problem you solve once. It needs the same ongoing maintenance you already give your SEO.
What this means if you make content
For creators: conversational Siri with memory of messages and photos means more people will ask their phone a question instead of opening an app or a search engine to find your content. If you publish how-to, product comparison, or explainer content, structure it so an assistant can lift a clean, correct, attributable answer out of it, a direct question-and-answer format works better here than a narrative lead-in. That’s a different discipline than writing for a feed or a search snippet, and the creators who build the habit early are the ones an assistant is most likely to quote once this becomes a bigger share of how people search. It’s also worth testing your own content directly: ask Gemini, ChatGPT, and Claude a question your content answers, and see whether any of them surface your work, credit you, or even get the answer right. If none of them do, that’s the gap to close before the assistant layer becomes a meaningful traffic source rather than a curiosity.
Common questions
Is Siri now literally Google’s product? No. Apple still owns and controls the Siri product, interface, and the data it accesses on your device. What’s reportedly changed is which company’s model handles the reasoning behind the scenes, similar to how many apps use cloud AI providers without being owned by them.
Can I actually switch Siri to Claude or ChatGPT right now? Not yet, as of this writing. The code found in iOS 27 shows Apple built the underlying architecture to support that kind of swap, but Apple has not made a public announcement enabling it for users.
Do I need iPhone 17 Pro to get the new Siri? No, the redesigned interface and conversational ability work on iPhone 11 and later. Only Apple’s most advanced on-device model is limited to iPhone Air and iPhone 17 Pro and later devices.
Does this affect my SEO rankings on Google Search directly? Not directly, Siri and Google Search are separate systems even when they share an underlying model. But the broader trend they’re both part of, AI systems answering questions directly instead of showing a list of links, applies to both, and the same GEO fundamentals help with each.
Why would Apple let Gemini power a core iPhone feature when Apple and Google compete directly on hardware and services? Because the two deals solve different problems for Apple. Google already pays Apple billions annually to stay the default search engine, a deal about distribution. This is reportedly a separate, technical licensing arrangement about capability, Apple needed a model that worked now, and building one in-house on the original timeline wasn’t happening. Companies license competitors’ technology into core products more often than most users assume; it’s a trade-off between speed and control, and Apple chose speed.
Quick checklist
- Update to iOS 27 and test the new Siri against real questions about your own brand or product.
- Ask Gemini, ChatGPT, and Claude directly what they know about you, and fix the gaps.
- Rewrite your FAQ and comparison pages in direct question-and-answer format, not just marketing copy.
- Watch for Apple’s official announcement on third-party model support in Siri. The code is already there; the public rollout isn’t, yet.
None of this requires action today beyond the audit. It does mean the assistant you’re optimizing for on someone’s iPhone isn’t a single, stable target anymore, and treating it as infrastructure rather than a one-time project is the difference between being ready for the next change and reacting to it. The 18 months Apple spent getting here is also a useful reminder for anyone planning around AI features generally: announced timelines for frontier AI capability, from any company, are consistently optimistic. Build your own roadmap with slack for the same kind of slip.
Sources: Apple Newsroom, September 2026; AppleInsider, September 14, 2026; TechCrunch, February 11, 2026, on the Gemini partnership and delay timeline.
Related reading
- Meta Muse Explained: What Meta’s Personal AI Agent Means for Marketers and Creators
- AI News for Marketers, Sep 11 2026: UMG x ElevenLabs, Suno v6
- Prompt Engineering for Marketers: 9 Hacks That Work in 2026
