iOS 27 Siri Third-Party AI Assistants Explained: Apple's Strategy

Apple is turning Siri into a platform. According to Bloomberg's Mark Gurman, confirmed by Reuters this week, iOS 27 will let third-party AI assistants including Claude, Gemini, Copilot, Grok, and Perplexity plug directly into Siri through a new Extensions system, ending the exclusive arrangement with ChatGPT that has been in place since iOS 18. The announcement is expected at WWDC on June 8.

The surface story is user choice. The deeper story is distribution. Apple's installed base runs to roughly two billion active devices, according to AOL's analysis published this week, and the company already takes a 30% cut on subscriptions processed through the App Store. Opening Siri to every major AI provider means Apple becomes the entry point for an industry it has largely watched from the sidelines.

This is less a concession to openness than a replication of Apple's oldest strategic move: own the front door, set the rules, and let others compete for what's behind it. The question isn't whether Apple is opening up. It's how open "open" actually turns out to be.

How iOS 27 Siri third-party AI assistants will work

The current ChatGPT integration is a simple handoff: when Siri can't answer a question, it offers to pass the query to ChatGPT with user permission. Siri's role in that exchange is limited. Extensions generalize and deepen that model.

Under iOS 27, users will manage which AI services connect to Siri through a toggle menu in the Apple Intelligence and Siri section of Settings. Language discovered inside pre-release builds reads: "Extensions allow agents from installed apps to work with Siri, the Siri app and other features on your devices," 9to5Mac reported this week. Apple will direct users to a dedicated App Store section to download additional providers, per the same report.

The feature ships across iOS 27, iPadOS 27, and macOS 27. MacRumors confirmed, citing Bloomberg's reporting, that AI companies will need to update their apps to support the new API before integration becomes functional, meaning the Extensions marketplace won't be fully populated on day one.

Importantly, Reuters reported this week that users may be able to choose which AI service handles each individual request, not just set a single global default. If that holds, Extensions isn't a preference toggle but a per-task routing layer: ask Siri to summarize an email with one AI, draft the reply with another. That distinction matters for both user experience and competition, since it changes whether providers are fighting for a default slot or for request-by-request preference.

What this means in practice for users and developers

For ordinary iPhone users, the practical shape of this is more complicated than the Settings toggle makes it look.

Getting the best models will likely require separate app installs and paid subscriptions. An iPhone user with Claude or the Gemini app installed can route Siri queries to those services; without the app, there's nothing to toggle. Apple will direct users to a dedicated App Store section to find and download providers, per 9to5Mac's reporting, which means discovery runs through Apple's storefront from the start.

There's also the latency question. Apple's Foundation Models framework, documented at WWDC 2025, handles on-device processing for routine tasks like summarization and drafting private, offline, and fast. Extensions-based providers route through the cloud. For many queries that difference won't be noticeable; for real-time or complex tasks, users may feel it.

Per-task routing, if Apple delivers on it, could be genuinely useful pick a specialized model for coding, a different one for creative writing. But it could just as easily become friction: too many options, inconsistent behavior across providers, and consent prompts on every handoff. The design of the routing layer matters enormously, and Apple hasn't shown it yet.

For developers, the integration requires updating apps to support the new Extensions API before the feature goes live, per MacRumors. Smaller or newer AI providers that can't move fast enough, or that find Apple's terms unfavorable, simply won't be there when iOS 27 ships.

One structural advantage is already baked in. Apple has said Gemini will underpin the more capable Siri experience it plans to ship in 2026, which gives Google's models a privileged position even before Extensions arrive, according to reporting from both The Verge and AOL's analysis. That deal is non-exclusive, but Extensions-based providers are starting one layer removed from where the defaults actually sit.

The real strategy: Siri as the AI distribution layer

Apple's willingness to open Siri is easier to understand once you accept that this isn't primarily an AI strategy. It's a distribution strategy, and a corrective one.

The past year was rough for Apple's AI credibility. The Verge catalogued the damage in January: Apple Intelligence summaries that got facts wrong, television ads promoting Siri features that still hadn't shipped, and the replacement of longtime AI chief John Giannandrea. Bloomberg Law noted this week that the Siri overhaul is explicitly aimed at reversing Apple's lag in AI relative to competitors. Opening Siri to third-party models is the fastest route to a useful assistant without waiting for homegrown AI to catch up.

That repair job has a revenue angle. Bloomberg's reporting, relayed by MacRumors and confirmed by Reuters, makes clear the Extensions model is expected to generate new App Store revenue. Apple's standard 30% commission on App Store subscriptions applies to AI services purchased through that channel. Users directed from Siri's Extensions menu toward Claude or Perplexity are being funneled through a storefront Apple controls and monetizes.

Apple's likely play is straightforward: spend less on frontier models, then monetize distribution if those models sell through its platforms. Amazon, Microsoft, Alphabet, and Meta collectively guided toward roughly $700 billion in AI-related capital expenditures for 2026, while Apple's fiscal 2025 spending came in at $12.7 billion, down 19% year-over-year, with projections of roughly $14 billion this year, per AOL. Apple isn't trying to win the model race. It's trying to own the distribution layer for whoever does.

What limits this: defaults, access terms, and the antitrust echo

The tension running through the Extensions model is that Apple is simultaneously the platform host and the most powerful competitor on it. That tension has at least three pressure points worth watching.

Who actually gets in. Access to Siri Extensions requires meeting Apple's standards on privacy, safety, and payment processing, per AOL. Those criteria aren't yet public. If the bar is high or commercial terms unfavorable, smaller and newer AI providers may opt out, leaving users with a Settings menu featuring the same three or four dominant names. The Extensions model looks open by design; it can still be narrow in practice.

Gemini's default advantage. The conversational Siri overhaul Apple plans to ship in 2026 is built on a Google Gemini backend, per Thurrott's reporting and AOL's analysis. On-device processing via Apple's Foundation Models framework handles simpler tasks privately and offline. Complex queries route to cloud models, and the cloud model anchoring native Siri is Gemini. Apple and Google stated jointly that Apple Intelligence "will continue to run on Apple devices and Private Cloud Compute, maintaining Apple's industry-leading privacy standards," per The Verge, with direct data sharing requiring user permission. The architecture is coherent. But third-party Extensions providers compete against a model that's already inside the walls.

The antitrust echo. Bloomberg Law's analysis from two months ago draws a direct parallel between the Extensions structure and the Google-Apple search deal, under which Google paid up to $20 billion annually for default placement on Apple devices. A federal judge concluded in 2024 that those arrangements were anticompetitive because default status alone produced foreclosure effects, even when alternatives were technically available. The Extensions model formally ends exclusivity, which weakens any such argument. But Apple retains control over who qualifies, which provider anchors native Siri, and what subscription fees apply.

MacRumors also notes that xAI previously sued Apple and OpenAI over the ChatGPT exclusivity arrangement, alleging anti-competitive conduct. Extensions resolves that specific complaint. The broader structural question whether Apple's gatekeeper role in AI distribution is itself a problem will outlast the June announcement.

What June 8 will reveal

The WWDC keynote will confirm or complicate everything above. Three specifics will determine whether this is a genuine platform opening or a well-designed press release.

The terms of access for third-party providers aren't yet public, and they're the whole game. If major AI labs accept Apple's commission structure, the Extensions marketplace becomes real. If they don't, providers will look for ways to steer users toward direct web billing, as some have already done in other App Store contexts, and Apple's revenue thesis weakens considerably.

Second, Apple needs to clarify how defaults work when multiple Extensions are active and whether users can route by task or only set a global preference. Third, the native Siri overhaul itself, including the standalone Siri app Thurrott reported is in testing, needs to actually ship. As Morningstar analyst William Kerwin told The Verge in January: "The long and short of it is that they over-promised back in the summer of 2024, and they under-delivered, still, now, what they promised." User trust in Apple's AI commitments is not something that recovers with a press release.

If the terms favor Apple heavily and defaults consistently route to Gemini, iOS 27 Siri third-party AI assistants will exist in name only more choice on paper, with Apple still owning the economics and controlling the floor. If providers accept the terms and Apple genuinely allows per-task routing, it could be the most consequential change to Siri since its launch. Those two outcomes are not equally likely, and June 8 will make clear which one Apple is actually building toward.