Apple launches Siri AI: What it means for Agentic Commerce

Apple has rebuilt Siri from the ground up. At WWDC 2026 the company unveiled Siri AI, a fully conversational assistant that holds multi-turn conversations, draws on real-time world knowledge, and can take action across the apps and data on your device. After two delayed attempts, the assistant Apple promised in 2024 is finally shipping, and how it was built has real consequences for how brands get found and recommended.

The all-new Siri

The headline features:

1. A dedicated Siri app. Siri now has its own home, with conversation history that syncs across devices through iCloud, so you can pick up past interactions where you left off.
   
   

2. It pulls from all your data across the device. Siri can reference information spread across your apps, for example reading a flight time from Mail, cross-referencing it against Calendar, and drafting a message without you leaving the conversation.
   
   

3. It has world knowledge. Siri now answers from real-time, up-to-date knowledge of the wider world, and can search the web, summarise, and generate content.
   
   

4. Notify me. Siri can watch for things on your behalf and alert you, for example when a product comes back in stock.
   
   

5. Agentic features. Siri can take action on your device, executing multi-step tasks across apps rather than just answering questions.
   
   

The thread tying these together, and Apple's real bet, is personal context. Siri can fold what is already on your device, your mail, messages, calendar, and app activity, into its answers. That is the capability no standalone chatbot can easily match, because none of them sit on the device that holds your life.

How it works

Three companies sit behind the new Siri, each owning a different layer.

Apple builds everything you actually see and touch. That means the Siri experience itself, plus the "traffic controller" that takes your request, works out what you are asking, and decides whether your phone can answer on its own or needs to send the request to the cloud. Apple also sets the privacy rules everything else follows, which is what lets it use your personal data without handing it to anyone else.

Google provides the brains. The AI models that understand language and generate answers are built on Google's Gemini technology, co-developed with Apple and rebadged as Apple's own. There is no Google app, no Google branding, and none of Google's own assistant. Apple licenses a custom version of the Gemini 3 model for integration into Siri across its ecosystem of over 2 billion active devices.

Nvidia provides the muscle. When a question is too big for your phone to handle alone, it goes to powerful computers in the cloud running on Nvidia's hardware. Apple does not own those machines, so it wraps them in its own security to meet its privacy standards.

Implications for agentic commerce

Conversational shopping used to take a deliberate act: open ChatGPT, Gemini, or Amazon, and start a conversation. Siri AI quietly deletes that step. A capable agent now lives one swipe from the lock screen on 2 billion devices, so the distance between "I want something" and "ask an AI to find it" has all but closed.

The bigger shift is that Siri starts each conversation already knowing something about you. It can draw on your mail, your calendar, your past activity, turning a generic query into a personal one. That is the part rivals will struggle to copy, and it is what tends to turn a novelty into a habit. An entry point this convenient, and this personal, is exactly the kind that pulls people in fast. Brands should expect conversational shopping on Apple devices to spike, not trickle.

So the question becomes: when Siri recommends a product, where does that recommendation actually come from? Not Apple. Siri's intelligence is Gemini, and Gemini's shopping answers are built on Google's Shopping Graph, the product knowledge layer fed primarily by Google Merchant Center and topped up with content Google crawls across the open web. Apple owns the interface, the privacy guardrails, and the personal context, but the product knowledge driving commerce answers is Google's.

Three consequences follow:

1. A wave of conversational queries lands on Google's commerce plumbing. Millions of iPhone shopping requests are about to route through Gemini rather than typed search or app browsing.
   
   

2. Personal context picks the winner, Gemini sets the shortlist. Siri uses what it knows about you to filter and rank, but it can only choose from the products Gemini already knows. Being complete and accurate in the feed gets you onto the shortlist, being the obvious fit for that shopper wins the recommendation.
   
   

3. Acting on your behalf raises the stakes on data. Siri does not just suggest now, it drafts, notifies, and completes tasks. Once an assistant acts for a shopper, a wrong price or a stale stock status is no longer a missed impression, it is a wrong action taken. Feed accuracy becomes a transaction-integrity issue, not a discoverability one.

What brands should do to prepare

If Siri-driven shopping is about to climb and the answers come from Gemini, preparing for Siri means optimising hard for Google's models, while making your products specific enough for personal context to latch onto.

Two inputs do most of the work: your Google Merchant Center feed, and the accessibility and machine readability of your web content.

Start with the feed. Merchant Center is the primary input into Google's Shopping Graph, and a complete, approved feed means your products are represented accurately in the layer Gemini and Google Shopping draw from. Audit it for completeness across title, description, product type, Google product category, GTIN, availability, price, variants, and the category-relevant attributes shoppers actually ask about. Attributes describing who and what a product is for, use case, audience, occasion, compatibility, matter most here, because those are exactly what Siri matches against what it knows about the shopper.

Then make your owned content readable by machines, not just people. Make sure important pages are crawlable, allow legitimate AI and search crawlers where appropriate, implement product and article schema consistently, and avoid hiding core product content behind JavaScript that may not be reliably processed. Schema is no guarantee on its own, but it helps these systems understand page purpose, product details, and the relationships between them.

The practical work, drawn from the Gemini section of our recently announced 5C Agentic Commerce framework:

1. Map what shoppers will ask Siri. Siri fields full, spoken questions, not keyword fragments, and many arrive with context built. Build a picture of how customers actually phrase requests: pull keyword data filtered for who, what, where, when, why and how, mine People Also Ask boxes, and read how people ask about your products on Reddit and Quora. Watch the contextual cues, occasion, recipient, budget, since those are the hooks personal context will use.
   
   

2. Check whether your content answers them. Take the questions that matter most in your category and ask where Siri would find the answer: in your feed, on your product pages, in your FAQs, or nowhere. Compare what shoppers ask against what your brand.com pages and feed actually contain. Any high-intent question that goes unanswered is the first gap to close.
   
   

3. Fix the inputs you control. Four levers do the work. Write product titles and descriptions that answer real questions in plain, specific language rather than marketing copy. Use FAQ content for the comparison and use-case questions a PDP cannot fit. Add product and article schema so Google parses your pages cleanly. And complete your Merchant Center attributes, since those are the signals Google uses to classify and recommend, and what Siri inherits.
   
   

4. Show up in the sources Siri's model trusts. Gemini does not answer from your pages alone, it synthesises across retailer PDPs, brand sites, earned media, and UGC. In Azoma's analysis of shopping agent citations, Gemini's sources broke down roughly to 41% retailer, 37% earned media, 15% brand.com, and 7% UGC. Retailer listings and earned media carry most of the weight, so getting recommended by Siri is a job for PR and ecommerce together, not your content team alone.
   
   

5. Track how Siri recommends you. Doing the work is one thing, knowing it worked is another, and personal context makes this harder than ever. The same query can return different products for different shoppers, so there is no single ranking to check. Monitor how often Siri surfaces your products, against which queries, and where you sit versus competitors. That tells you which gaps to close next and whether your feed and content changes are translating into real recommendations.
   
   

The bottom line

Siri AI is not a new product knowledge graph for brands to learn. It is a very large new front door onto Google's existing one, opening just as conversational shopping is about to get a lot more common. Brands that have already done the Gemini work, a complete Merchant Center feed, crawlable and structured content, and a consistent signal across the sources AI cites, are well placed to surface in Siri's answers.

But Siri has its own nuances. The personal context layer is Apple's alone, and it sits on top of Gemini in ways plain Gemini optimisation will not fully capture.

That is where Azoma comes in. We help brands optimise for both the Gemini models underneath and the Apple-specific layer on top, and track how you're showing up. Since personal context means the same query returns different products for different people, we use shopper personas to measure your share of voice across shopper types, not just on average, and help you make subsequent optimisations at scale.

👉 Want to know how Siri AI is recommending your products? Get in touch today to learn more.