Agentic commerce has a new measurement problem. NIQ and Similarweb said September 2 that they are building a system to track whether products appear in AI-driven discovery, whether product content is ready for agents, and whether those interactions lead to traffic, conversion, and sales.
For enterprise retailers, the move reinforces a shift already visible across Adobe, OpenAI, and Shopify: checkout is only the end of the stack. AI systems first need accurate product attributes, pricing, inventory, availability, and other catalog data before they can recommend or transact on a product.
Product data moves ahead of payment integration
NIQ’s September 2 announcement says the joint effort will measure five areas: consumer intent, agentic shelf visibility, product-content readiness, AI-driven traffic, and AI-driven conversion. An initial version is planned for Q4 2026, starting with a limited set of categories and markets.
Adobe made the infrastructure case a month earlier. Its August 7 Adobe Commerce release added Catalog Agent, which exposes structured product names, attributes, specifications, variants, pricing, availability, and product relationships to AI-powered discovery systems.
OpenAI also expanded the Agentic Commerce Protocol in March to support product feeds, promotions, and richer product discovery inside ChatGPT. The company said its first Instant Checkout implementation did not provide enough merchant flexibility, so it shifted more effort toward discovery while allowing merchants to use their own checkout experiences.
Shopify has formalized a similar split. It positions Catalog API as UCP’s product-discovery component, turning merchant catalogs into structured, queryable data for AI systems. Shopify says AI searches using Catalog data convert at twice the rate of searches relying on scraped data, although that is the company’s own measurement.
The market is also pushing AI shopping beyond recommendations. Amazon’s Alexa for Shopping can compare products, track prices, and help manage purchases, while Visa and OpenAI are working on agent-payment capabilities for transactions that require authorization and trust controls.
Current OpenAI shopping documentation says Instant Checkout can still appear for some eligible products and merchants. Walmart’s March test suggests keeping checkout inside an AI interface does not automatically improve conversion: purchases completed in ChatGPT converted at roughly one-third the rate of purchases completed after shoppers clicked through to Walmart.com. Walmart had made about 200,000 products available but did not publish the underlying dataset, according to reporting on the test.
Agent access also remains unsettled. An August court ruling allowed Perplexity’s shopping agent to remain on Amazon while litigation continues, showing that merchant control over outside agents is still being tested.
What eWeek found: Discovery is standardizing faster than checkout
Across the primary sources, product discovery is emerging as a distinct infrastructure layer before transaction execution. Adobe exposes catalog data to AI systems, OpenAI extended ACP into discovery, Shopify separates catalog discovery from broader UCP commerce functions, and NIQ now plans to measure product-content readiness alongside AI-driven conversion.
Checkout remains less uniform. OpenAI still supports Instant Checkout for some eligible products, Shopify routes ChatGPT purchases through merchant-controlled checkout, and Walmart reported weaker conversion for in-ChatGPT purchases than for click-through transactions.
For enterprise architecture teams, catalog quality, inventory accuracy, pricing, and attribution are more portable investments than custom payment work tied to one protocol or agent. Those data foundations can support multiple AI channels while checkout models, authorization rules, and merchant controls continue to evolve.
Read more: For a broader view of the controls and deployment risks around autonomous systems, see the agentic AI guide to tools and protocols.
Want to learn more AI tips, tricks, and prompting techniques? eWeek readers get free 7-day access to The Neuron Academy, our practical learning platform designed to help professionals use AI more confidently at work. Browse all lessons →