AI Agents: E‑com Blindspots

Vercel’s server‑log analysis logged GPTBot at 569 million fetches and Anthropic’s Claude at 370 million in a single month, and reported that the major AI crawlers in its dataset do not execute JavaScript when fetching pages. (vercel.com) Adobe Analytics recorded a 1,300% jump in generative‑AI referrals to U.S. retail sites during Nov–Dec 2024 and reported that AI‑referred visitors browse 12% more pages per visit and have a 23% lower bounce rate vs. non‑AI traffic. (blog.adobe.com) Google unveiled the Universal Commerce Protocol (UCP) on January 11, 2026 as an open standard to let AI agents move from discovery to authenticated order flows across merchant systems. (developers.googleblog.com) Adobe said on February 18, 2026 that Adobe Commerce will support both Google’s UCP and OpenAI/Stripe’s Agentic Commerce Protocol (ACP) to make product catalogs, pricing and inventory machine‑readable for AI shopping surfaces. (digitalcommerce360.com) Google’s Product Variant structured‑data guidance documents the ProductGroup + Product pattern and the hasVariant/variesBy/productGroupID properties as the sanctioned method to expose variant relationships, pricing and shared attributes to automated systems. (developers.google.com) Generative‑crawler audits show a multi‑stage pipeline—fetch, render, extract, segment, embed—so pages that return partial or client‑only content become fragmented or mis‑embedded in agent datasets. (ranktracker.com) E‑commerce operational guides and vendor briefs now emphasize machine‑readable catalogs, consistent feeds, stable APIs for inventory/pricing, and explicit transactional rules so agents can compare, authorize and complete purchases without human navigation. (almcorp.com)