If a shopper asks ChatGPT or Claude to order dinner through an Android app today, the assistant can talk about it but can’t actually reach in and finish the order β€” only Google’s own Gemini has that system-level access. A binding EU ruling issued this month means that gap closes on a fixed timeline, and it changes who’s allowed to attempt a purchase inside a merchant’s Android app, not just how search rankings get shared.

What the European Commission ordered

On July 16, 2026, the European Commission adopted two binding “specification measures” against Google/Alphabet under the Digital Markets Act’s interoperability rules (European Commission DMA developer portal):

  • Android AI interoperability. Google must give rival AI assistants free and effective access to eleven defined Android system features under DMA Article 6(7) β€” the same features Gemini currently uses exclusively to respond to voice wake words, read on-screen context and reach into other apps to complete tasks like sending a message, sharing a photo, or ordering food (TECHi, “EU Opens 11 Android Features to Rival AI Assistants”; The Hacker News, “EU Orders Google to Open Android Mic, Camera and Screen to Rival AI Assistants”). Google must ship this with Android 18 by August 1, 2027; support for concurrent wake words (letting a phone answer to more than one assistant’s trigger phrase) follows with Android 19 by August 2028.
  • Search data sharing. From January 2027, Google must share anonymized Search ranking, query, click and view data with vetted third-party search engines and AI chatbots on FRAND terms.

Neither decision carries a fine on its own, but non-compliance exposes Google to the DMA’s standard penalty of up to 10% of global annual turnover. Google disputes both rulings: President of Global Affairs Kent Walker said the Commission “discount[s] extensive evidence of user harm” and that sharing search data risks exposing users’ queries “without adequate anonymization” (Google, “The DMA should not undercut security & privacy for Europeans”).

Why this is a merchant problem, not just a Google-vs-rivals fight

The interoperability order is the part that matters for e-commerce. Today, an Android app’s in-app purchase flow β€” order a meal, rebook a ride, buy a product β€” already works with Gemini because Google built that integration at the OS level, outside the app developer’s control. Every other assistant is confined to its own app sandbox and can’t reach in the same way.

From mid-2027, that privileged access stops being Gemini-only. ChatGPT, Claude and other qualifying assistants gain the right to the same eleven Android hooks β€” meaning a wider, less predictable set of agents will start attempting to execute purchase actions inside merchant apps that were never built or tested against anything but Gemini’s specific implementation. An app whose “buy” flow only “works” because Google special-cased it is not the same as an app whose purchase action is genuinely well-structured for any agent that calls it.

That’s the same problem this site has tracked at the browser layer: WebMCP lets a site register a checkout tool an in-browser agent can call directly instead of guessing at a page’s DOM (our WebMCP coverage). The Android ruling pushes the identical question up a layer β€” onto native apps, where the “tool” isn’t a JavaScript function but whatever structured action (Android App Actions, deep links, intents) the app exposes, or doesn’t.

What to check before mid-2027

  • Don’t assume “it works with Gemini” means it works with agents generally. Confirm whether your Android app’s purchase/ordering actions are exposed through a structured, documented interface, or only function because of a Google-specific integration.
  • Expect more agents, not just more traffic. The assistants named as qualifying under the ruling β€” ChatGPT, Claude and others β€” are the same ones already shopping on the web, so the in-app channel is about to carry the same agentic-commerce traffic your site already gets.
  • Re-check after August 2027. The rollout date is fixed by the Commission’s order, not by vendor roadmaps, so this is one of the few agentic-commerce deadlines on a firm regulatory clock.

Whether an agent can find your Android app was never the hard part β€” Google’s own app store search already handles that. Whether it can complete a purchase once it has system-level access is the exact transaction-layer question an AgentReady scan is built to check.

Sources