No blockbuster launches today. None of the usual movers — OpenAI, Google AI Mode, Shopify, Stripe, PayPal, Walmart, Instacart — shipped new ad inventory or an agentic-checkout product. The signal came from two other places: Europe's highest court upheld a roughly 4.1 billion euro antitrust fine against Google over Android, and a cluster of operators showed AI search and agentic commerce hardening into real business plumbing rather than stage demos.
Key Takeaways
- Europe's top court rejected Google's appeal and upheld the ~4.1B euro Android antitrust fine, sharpening scrutiny of default search entry points and AI answer layers.
- Agentic commerce is moving from chat recommendations toward interfaces, preferences, after-sales and liability — the parts that decide whether an outside agent can actually complete a purchase.
- B2B buyers now ask LLMs directly for the right product fit, so dealer reviews, third-party tests and forums feed the model's answer, not just your own site.
- Structured preference data, like Instacart's Preference Picker, is becoming the fuel for AI shopping assistants, shifting recommendation from keyword match to intent plus quality preference.
- Adland is resetting expectations: the near-term AI payoff is system and data efficiency, not fully automated creative.
Europe's top court hands Google a final antitrust loss
On July 2 the EU's highest court rejected the Google/Alphabet appeal and upheld a roughly 4.1 billion euro Android antitrust penalty, centered on Search and Chrome pre-installation and Google's search advantage on Android devices, as reported by AP, WSJ and MarketWatch. This is not a new AI Mode case, but it lands on the same nerve: who gets to own the default answer layer.
For globalizing brands the read-through is distribution risk. As AI search becomes the default entry point on more devices, the legality of pre-installed defaults, search advantage and ad monetization will stay under a microscope. If regulators keep prying open default entry points, visibility inside AI answers becomes less about buying a slot and more about being the entity the model independently trusts.
Amazon, Swiggy and Zepto show three roads into agentic commerce
A roundup in The Economic Times mapped three distinct approaches: Amazon leaning on personalized memory and Rufus, Zepto running a multi-agent after-sales system called Zap, and Swiggy exposing MCP/API so external AI assistants can place orders, buy groceries and book tables. Different bets, same direction of travel.
The competition is no longer only about front-end recommendation. It is about whether an outside agent can reliably understand, call and complete a transaction against your catalog. That means product data, inventory, refunds, preferences and order status all need agent-readable interfaces. Brands that leave those as human-only screens will be invisible to the agents doing the buying.
