Siri AI Developer Preview: App Intents and Limits

Apple has opened a Developer Preview that lets any app expose its content and actions to Siri AI through App Intents. Announced on November 3, 2026, the preview is available with the iOS 27.3, iPadOS 27.3, macOS 27.3 and visionOS 27.3 betas, giving developers an early path to build voice-driven app experiences before customer availability expands.

The important change is not simply that more apps can answer a voice request. Apple says Siri AI can use natural language, personal context and onscreen awareness to find app content and run actions, while a new skills mechanism can combine several actions into one experience. That could make tasks such as checking a delivery, booking a reservation or placing an order feel more conversational, but only when the app has modeled its data and actions clearly.

This remains a developer beta. Apple has not announced a precise public launch date for third-party integrations, App Store submissions are not scheduled to open until spring 2027, and the underlying Siri AI service carries device, language, region, age and daily-usage restrictions. Developers should treat the preview as an integration and testing window rather than a finished distribution channel.

Siri AI developer preview at a glance

Availability: the Developer Preview is included in the 27.3 beta cycle for iPhone, iPad, Mac and Apple Vision Pro; Apple also lists Siri AI itself as a beta on watchOS 27.3.
Integration foundation: apps use App Intents, app entities and app enums to describe actions and content to the system.
New capabilities: natural-language requests, personal-context understanding, onscreen awareness and access to entities indexed in Spotlight.
Complex workflows: Apple is introducing skills that can combine multiple app actions into one Siri AI experience.
Distribution timing: Apple says customer shipping is still a few months away and App Store submissions will open in spring 2027.
Current languages: English, French, Japanese, Korean, Portuguese and Spanish, with more languages planned.
Important limits: compatible Apple Intelligence hardware and software are required; availability varies by region; users under 13 are excluded; some server-backed features have daily limits.

Those conditions make the preview most useful for teams that already have a clear App Intents model or can build one without changing their core app architecture. A voice assistant cannot safely infer every private object, state change or business rule. The developer still defines which entities are discoverable, which actions can run and how results return to the user.

How Siri AI connects app content and actions

App Intents is the bridge between an app and Apple Intelligence. An app defines an intent for an action, entities for the objects that action uses, and enums for bounded choices. Apple Intelligence can then match a conversational request to those structures instead of trying to operate an app by guessing at interface coordinates or scraping arbitrary screens.

Apple’s new documentation identifies several layers that affect whether an integration works well:

Spotlight indexing makes app entities findable even when a person describes an item vaguely rather than using its exact name.
Transferable types let compatible content move between apps, which matters when a request starts in one app and finishes in another.
Schemas give actions and entities a known structure so Siri AI can match everyday language to the correct operation and parameters.
Onscreen associations connect visible interface elements with app entities, allowing references such as “this photo” or “that event.”
Action and content donations provide behavioral cues that can help the system disambiguate requests and suggest repeated tasks.

This is a structured integration, not an unrestricted grant of control. If an app has not exposed a supported action, indexed an entity or associated onscreen content with a model the system understands, Siri AI may be unable to complete the request. Developers should test the exact vocabulary, parameters, failure states and confirmation flow for each action they publish.

What Siri AI skills add beyond a single App Intent

Apple uses the term skills for experiences that teach Siri AI how to combine several actions. A restaurant app could expose separate actions for searching availability, selecting a time and booking; a delivery app could expose tracking, address updates and notification controls. A skill can help organize those pieces into a coherent conversational sequence instead of forcing a person to issue each command separately.

Apple has not yet published all terms and usage policies for Siri AI-enabled experiences. That gap matters for commerce, account changes, health data and other high-impact actions. A developer can prototype now, but should avoid designing a launch plan around assumptions about final review rules, permitted transaction types or background behavior until Apple documents them.

The safest design is to keep state-changing actions explicit. Siri AI should summarize what will happen, surface the account or item being affected and request confirmation when a mistake would be difficult to reverse. Read-only lookups, such as checking delivery status, generally need less friction than canceling an order, changing a reservation or deleting content.

Spotlight indexing becomes more important

Apple says Siri AI can now surface any entity in the Spotlight semantic index, rather than only entities that fit an existing schema. That broadens discovery for apps with specialized content, but it also raises a practical data-modeling question: which objects should be searchable outside the app, and what metadata is necessary for a useful match?

Developers should index the minimum information needed to identify an entity and complete an intended action. Internal notes, hidden fields, expired objects and confidential records should not become discoverable merely because an index can store them. Search relevance also benefits from clean titles, alternative names and current state; stale or duplicated entities make conversational results harder to trust.

A useful test set should include vague descriptions, similar entity names, outdated items, multiple accounts and cases where no result is safer than a weak match. The assistant’s confidence is not a substitute for app-side validation. Before an action runs, the app still needs to enforce authorization, availability, inventory, subscription and other business rules.

Device, language, region and account limits

The Developer Preview does not make Siri AI available on every Apple device or in every market. Apple lists several gates that developers and testers should check before reporting a bug or promising a feature:

The device must support Apple Intelligence and run a compatible 27.3 beta operating system.
The device language must be one Siri AI currently supports: English, French, Japanese, Korean, Portuguese or Spanish.
Some capabilities may differ by language or region even when the main Siri AI interface is available.
Siri AI is currently unavailable in the European Union on iOS, iPadOS and watchOS.
Apple says Siri AI is not available to users under 13.
Server-backed Apple Intelligence features, including Siri AI, can be subject to daily usage limits that vary with request complexity, demand and policy.

Apple also says expanded access to limited server-backed features will be available for a fee in the future. It has not published the consumer or developer pricing structure. An app should therefore degrade gracefully when Siri AI is unavailable or a usage limit is reached, keeping its normal interface and core functions usable without the assistant.

A practical developer checklist

Inventory the actions people genuinely want to complete by voice and remove actions that are too ambiguous or risky for conversational execution.
Define app entities and parameters with names, synonyms and constraints that reflect the app’s real data model.
Adopt relevant schemas where Apple provides them, and use custom entities only where the system schemas do not fit.
Index only appropriate entities in Spotlight, with deletion and refresh behavior that prevents stale results.
Associate visible content with entities so references to the current screen resolve predictably.
Test read-only, state-changing, destructive and transactional actions separately, including cancellation and recovery paths.
Verify permissions and authentication at execution time; do not assume that a successful Siri match grants access.
Test unsupported languages, EU accounts, under-13 accounts, offline behavior and exhausted usage limits.
Keep the standard app interface complete so users are never forced to rely on Siri AI for an essential action.

Teams should also log failures in a privacy-conscious way. A transcript can contain personal requests, account details or sensitive app data. Diagnostics should collect only what is necessary, redact secrets and avoid treating a conversational input as harmless telemetry.

What the preview means for users

For users, the promise is simpler access to app features without memorizing exact command phrases. A person might ask for a delivery in ordinary language, refer to the photo currently on screen or chain several related actions into one request. Because the integration is defined through App Intents, an app can return structured results and enforce its own rules instead of handing Siri AI unrestricted interface control.

The limits are equally important. A request may fail because the app has not implemented the necessary intent, an entity is not indexed, the device or language is unsupported, the account is in a restricted region or a daily service limit has been reached. Users should continue to review confirmations and use the app directly for sensitive or high-value actions when the assistant’s result is unclear.

An iPhone buying note

Siri AI requires an Apple Intelligence-enabled device. Readers comparing current hardware can review the Apple iPhone 18 Pro 256GB unlocked listing on Amazon (ASIN B0HJHVB95Z). Confirm the seller, storage, carrier compatibility, return policy and current price before purchase. The listing does not guarantee that every Siri AI feature is available in a buyer’s country, language or account.

Buying new hardware is not required for developers who already have a compatible test device. It is also not a shortcut around regional, language, age or usage restrictions. Teams should verify Apple’s current support matrix before purchasing equipment solely for this preview.

Affiliate disclosure: ReaderSpice may earn a commission from qualifying Amazon purchases made through the link above, at no additional cost to you.

Bottom line

Opening Siri AI to every developer is a meaningful expansion of Apple’s app integration model. App Intents, Spotlight entities, onscreen context and multi-action skills give developers a structured way to make app content and actions available through natural language. The best integrations will be narrow enough to remain predictable, but complete enough to handle real user goals.

For now, this is an early preview with unfinished distribution rules and significant availability limits. Developers should use the 27.3 beta cycle to model entities, define safe actions, test ambiguous requests and build graceful fallbacks. A public launch plan should wait for Apple’s final policies, App Store submission details and supported-device matrix.