Agentic Commerce Before Peak Season: Google Expands Its Toolkit — and Onsite Search Becomes the Control Point

Weeks before the most important quarter of the retail year, Google unveiled a package of agentic commerce updates on 16 September. The company made "AI performance insights" in Merchant Center generally available in Australia, Canada, India, New Zealand and the United States. The report shows retailers how their brand and products are discovered across AI Mode and AI Overviews, benchmarked against competitors. There is still no click metric, so the feature remains a platform-specific diagnostic rather than a complete picture of demand.

Alongside it, Google opened a US beta embedding its Business Agent directly into YouTube ads, letting viewers ask detailed product questions without leaving the video. The Universal Commerce Protocol (UCP) integration hub also gains cart transfer — moving a basket from Google to the merchant's own site — plus enhanced checkout-flow testing. Australia and Canada are expected early next year. Cart transfer answers retailers' loudest complaint about checkout-inside-AI: the customer relationship, login data and loyalty engagement stay with the store. Tapestry, owner of Coach and Kate Spade, already uses UCP-powered checkout across Search, AI Mode and the Gemini app.

The timing is deliberate. Adobe data shows traffic from AI sources to US retail sites grew 393% year over year in Q1 2026, after a 693% surge during the 2025 holiday season. Quality matters more than volume: by March 2026, AI-referred traffic converted 42% better than conventional channels — a complete reversal within twelve months, since the same traffic converted markedly worse a year earlier. Salesforce predicts 20% of 2026 holiday ecommerce traffic will originate from AI chat agents, and that one in three ecommerce sites will run a site-specific shopper agent by Cyber Week.

A PYMNTS Intelligence report published this month tempers the enthusiasm. AI has clearly won a place in product discovery — roughly 132 million US adults have bought a retail product with AI's help — but consumers still control the purchase, and Amazon captures a disproportionate share when research converts into a sale. Merchants expect agentic commerce to become a paid sales channel, yet many say they are unprepared to connect to AI platforms or agent networks. OpenAI's decision in March 2026 to retire Instant Checkout and pivot to "discover in the chat, transact on the merchant site" points in the same direction.

For onsite search vendors, the role is shifting. The search index is no longer just a front-end feature; it is the structured data source external agents draw on. Gartner framed the category accordingly in its Magic Quadrant for Search and Product Discovery, published 22 June 2026, citing natural language processing, intent detection, algorithmic merchandising and agentic integrations. Algolia was named a Leader for a third consecutive year, alongside Coveo, Constructor and Netcore Unbxd, among the twelve vendors evaluated. Product roadmaps reflect the same pressure: agent studios, MCP servers and feed enrichment aimed at AI visibility.

Europe shows the pattern in practice. Otto launched a voice- and text-based shopping assistant built on Google Gemini in March 2026, interpreting free-form requests across more than 18 million articles and refining them with follow-up questions. Zalando's assistant handled close to ten million queries in Q1 2026 alone, against roughly six million across all of 2025.

The practical takeaway for retailers is unglamorous but uncomfortable: clean feeds, machine-readable attributes for price, availability, shipping and returns, open crawler access, and a reliable first-party checkout. Skip that groundwork before peak season, and you simply will not appear in the agentic discovery layer.