Siri AI App Skills: What the iOS 27.3 Developer Preview Opens Up

Apple opened the Siri AI Developer Preview to all developers on November 3. Apps can expose custom actions and content through App Intents, while new app skills guide Siri through tasks involving several actions. For developers, the change expands what an assistant can actually do inside their product.

The important availability detail is easy to miss: this is a developer preview, not permission to ship these integrations to every customer today.

What changed in Siri AI

Apple's announcement covers the iOS 27.3, iPadOS 27.3, macOS 27.3 and visionOS 27.3 betas. Siri can now work with custom actions and content beyond Apple's predefined app schemas. It can also surface entities from Spotlight's semantic index without requiring them to fit an existing schema.

That gives developers more room to describe what their app actually does, rather than squeezing its functionality into categories Apple already understands.

App Intents remain the integration layer. Siri only gets the actions and content a developer explicitly exposes. Opening the preview does not give it unrestricted access to an app's internal state.

The preview documentation describes how context affects discovery. With an app in the foreground, an active Live Activity, or a continuing Siri conversation that mentions the app, users do not always need to repeat its name. For a background app, they need to name it in their request.

For a founder, that distinction matters. Someone already using a product can ask Siri to continue a task with less repetition. Reaching a user who is outside the product still depends on that user knowing which app they want.

How Siri app skills work

An individual tool intent handles an action. A skill supplies the context needed to combine actions: which tool is appropriate, how steps relate, what to avoid and what domain knowledge matters.

Apple's app skills documentation describes Markdown files packaged in a Skills.skillsbundle. Each skill has a name, a description and a body. Siri uses the description to decide whether the skill fits a request, then reads the body when it needs the guidance.

The format uses a subset of Agent Skills. It supports the name and description fields plus Markdown, but not accompanying scripts or reference files. The description has a 1,024-character limit and the body a 10,000-character limit.

Apple uses shopping as an example: finding items, adding them to a cart, choosing delivery and payment methods, reviewing the cart and confirming a purchase. That example shows why a list of individually callable functions is sometimes insufficient. An assistant also needs to understand the relationship between those functions.

There is an important limit here. App skills cannot enforce deterministic behavior. Siri does not simply execute every listed intent in order. Developers are giving a model guidance, so the quality of that guidance and the behavior of the underlying tools both matter.

Who can test the preview

Developers can build a development-signed app in Xcode and run it on a supported device, or distribute it to internal TestFlight testers. No additional entitlement or separate developer program is required for this integration.

External TestFlight testers and App Store users do not get the preview capabilities. Apple says customer availability starts in a few months, with App Store submissions opening in the spring. It has not given an exact public release date in this announcement.

The preview requires Apple Intelligence-compatible hardware. Siri AI also has region, language and age restrictions, including unavailability on iPhone, iPad and Apple Watch in the EU and for users under 13. Developers should distinguish those broader Siri restrictions from the four platforms listed for the custom-app preview.

Apple says terms and usage policies for these experiences will follow. Server-backed features have daily usage limits, with expanded paid access planned, but the announcement does not specify prices or fixed quotas.

The work moves into testing the interface

Apple includes a Siri Integration Advisor in Xcode to suggest improvements to App Intents, plus a Siri Evaluations framework for scripting requests and measuring how responses change as an app evolves. The APIs and Siri's behavior may change during the preview.

This makes the testing window useful even before distribution opens. A team can start with one narrow workflow and ask the same thing in different ways. Does Siri choose the right action? Does it understand the relevant entity? Does a vague request lead to a sensible next step?

Those are different tests from checking whether an API returns the correct response. The tool can work perfectly while the assistant selects the wrong tool or misunderstands the user's intent.

The practical opportunity is to make a product's useful actions understandable outside its own screens. The teams that use this preview well will leave with evidence about which requests work reliably, not just a new integration listed on a launch page.

I’m Paras. I work at the intersection of AI, engineering and distribution.

I help technical founders turn product updates and complex engineering work into clear stories developers and customers can understand and share.

Book a 30-minute call.