On June 8, 2026, minutes after the WWDC keynote in Cupertino, Apple seeded the iOS 27 developer beta and—buried in the release notes—posted a formal deprecation notice for SiriKit. The one-line summary: App Intents is now the only supported framework for exposing your app to the new, Gemini-powered Siri. If your app still drives Siri through SiriKit intents domains, the migration clock just started ticking.
To see why this matters, remember how we got here. SiriKit shipped in 2016 with a fixed menu of "intent domains"—messaging, payments, ride-booking, workouts—and if your use case did not fit one of Apple's pre-baked buckets, you were stuck. App Intents arrived in 2022 as the modern, code-first replacement: you declare an AppIntent struct, Apple indexes it, and it surfaces in Siri, Spotlight, Shortcuts, and Widgets. For four years the two frameworks coexisted and teams could procrastinate. June 8 ended the truce: Apple moved from "App Intents is preferred" to "SiriKit is deprecated," and tied the new Siri's assistant capabilities exclusively to App Intents.
Who is affected? Effectively every team shipping a serious iOS app—and for cross-platform shops, every Flutter or React Native app that wired a native Siri extension. The new Siri runs on Gemini under a multi-year deal reported at roughly US$1 billion per year, which means assistant traffic is about to get a lot more capable and a lot more used. Apple's guidance gives developers a window of roughly two to three years before SiriKit-dependent features stop working.
Below: the actual App Intents code you write, how it differs from a SiriKit intent, a migration-cost table, two things the keynote glossed over, and where a Flutter agency should place its bets.
Technical Details
A SiriKit integration meant an Intents app extension, an .intentdefinition file, and an INExtension handler living in a separate target. App Intents collapses all of that into a struct compiled into your main binary:
import AppIntents
struct BookSeatIntent: AppIntent {
static var title: LocalizedStringResource = "Book a seat"
static var description = IntentDescription("Reserve a seat for a class or event.")
@Parameter(title: "Class")
var session: SessionEntity
@Parameter(title: "Seats")
var seats: Int
func perform() async throws -> some IntentResult & ProvidesDialog {
let booking = try await BookingService.shared.reserve(session.id, seats: seats)
return .result(dialog: "Booked (seats) seat(s) for (session.name).")
}
}The shift is conceptual, not just syntactic. SiriKit asked, "which of Apple's domains does your feature resemble?" App Intents asks, "what verbs and entities does your app expose?" You model your domain (SessionEntity, BookSeatIntent) and the system—now backed by Gemini's language understanding—maps messy natural language onto your typed parameters. The same intent automatically becomes a Shortcut, a Spotlight result, and an interactive Widget with zero extra code.
Immediate Actions for Three Audiences
- Engineers: Install the iOS 27 beta and audit every target for
INExtensionand.intentdefinitionfiles. Port the highest-traffic intent to anAppIntentstruct first; verify it appears in Shortcuts before touching Siri. - Tech leads: Inventory which client apps still depend on SiriKit. Treat the 2–3 year window as real but not generous—schedule migrations into normal release cycles now, not as a panic in 2028.
- Founders / agencies: Package a fixed-scope "SiriKit → App Intents migration + assistant readiness audit" service. Clients with older iOS apps have a deadline they cannot ignore, which makes this an easy, time-boxed sale.
Comparison & Trade-offs
| Option | Pros | Cons | Migration cost |
|---|---|---|---|
| Stay on SiriKit | No work today; existing code keeps running short-term | Deprecated; cut off from Gemini Siri; dead in ~2–3 years | Deferred, but compounding |
| Migrate to App Intents | Future-proof; free Shortcuts/Spotlight/Widget surfaces; Gemini understanding | Re-modeling domain entities; new testing surface | ~3–10 dev-days per app depending on intent count |
| Drop Siri integration | Zero maintenance | Lose hands-free + assistant discovery entirely | Low, but a visibility loss |
What They Won't Tell You
- "Deprecated" is not "removed," but the moat is gone. Your SiriKit code runs for now, yet new assistant capabilities ship App-Intents-only. Standing still means quietly falling out of the assistant surface while competitors show up there.
- Gemini in the loop changes your testing burden. A probabilistic LLM mapping language to your parameters means you must test ambiguous phrasings and failure dialog—App Intents' typed parameters help, but they do not eliminate the new QA surface.
- Privacy and the Apple–Google relationship are a strategic risk. Siri now leans on a Google model (on Apple's Private Cloud Compute); a contract that large is also a dependency you do not control.
What Happens in the Next 3 Months
Expect Xcode 17 migration assistants that scaffold AppIntent structs from existing .intentdefinition files, a wave of WWDC session videos and sample code, and third-party libraries racing to wrap App Intents for Flutter/React Native bridges. Watch one signal: when Apple's own first-party sample apps drop their last SiriKit target—that is the unofficial "safe to fully commit" marker.
FAQ
Do I have to rewrite my whole app?
No. Only the parts that expose actions to Siri/Shortcuts. Most apps have a handful of intents; you re-model those as AppIntent structs and leave the rest of the app untouched.
How long until SiriKit actually stops working?
Apple's deprecation guidance points to roughly a 2–3 year window. Deprecated APIs keep running until a future removal; plan migrations into normal release cycles rather than waiting for a hard cutoff.
Does App Intents work for Flutter or React Native apps?
Yes, through a native platform channel/bridge. You write the AppIntent in Swift in the iOS runner and call into your Dart/JS logic. The native surface is unavoidable, but it is small and self-contained.
Will my Shortcuts break during migration?
Properly modeled App Intents power Shortcuts automatically. If you migrate carefully and keep intent names/identifiers stable where possible, user-built Shortcuts continue to function.
My Take (contrarian)
The popular read is "Apple finally cleaned up its assistant APIs." My read is different: the real story is that your app's discoverability is being re-platformed onto a language model you do not own. SiriKit's rigid domains were annoying, but they were deterministic. App Intents plus Gemini means a probabilistic layer now decides when your action surfaces to a user—and that is a new dependency dressed up as a developer convenience. For ScriptWalker the business implication is concrete: do not sell "App Intents migration" as a checkbox. Sell it as assistant visibility—model the client's domain entities well, write the dialog and disambiguation carefully, and treat the assistant as a new acquisition channel. The studios that win the next two years are the ones that move early and treat App Intents as product surface, not plumbing.