Structured product data becomes crucial as AI agents reshape e-commerce discovery

As AI-driven agents take on decision-making roles in online shopping, merchants must prioritise structured, machine-readable product information to remain visible and competitive in the evolving digital marketplace.

Devika Naik’s argument is straightforward: as commerce becomes more agent-driven, the decisive audience may no longer be a person scanning a page, but software acting for that person. The shift matters because product discovery has always depended on translating messy intent into structured information, and that task becomes more demanding when an agent is doing the interpretation rather than a shopper. In Naik’s telling, the real change is not just faster search or easier checkout, but a new layer of decision-making in which merchants must present products in a form that machines can understand and compare. According to related analysis from agentic commerce researchers, that requires data that is machine-readable, current and precise.

That is why structured product catalogues are becoming central to the discussion. A report from Agentic Commerce Report argues that schema.org Product markup gives AI systems a standard way to read core details such as name, description, price and availability. MongoDB has made a similar case, saying retail discovery is moving towards AI agents that research and compare products on behalf of consumers, which increases the value of clean, enriched catalogues and real-time inventory data. The practical point is simple: if a product cannot be parsed reliably, it is less likely to surface in an agent-led decision.

The implications go beyond metadata. FeedArc, in a guide on agentic commerce feeds, says merchants will need to supply trust signals such as seller policies, return windows, review summaries and price history, because agents will weigh those factors alongside the headline offer. The Universal Commerce Protocol blog makes the same broader case, arguing that an agent-ready catalogue should be retrievable, interpretable and actionable without scraping or guesswork. In other words, product pages built mainly for human persuasion may no longer be enough if the first evaluator is software.

That changes the role of advertising as well. Naik suggests that, in an agentic setting, merchants may need to provide evidence rather than merely capture attention. An agent can compare more options than a person typically would, and it is less affected by visual design or placement tricks. A model can instead assess compatibility, delivery promises, price and suitability in one pass. That places more value on truthful, structured information and less on the traditional storefront logic that has shaped digital marketing for years.

There are still unresolved questions around disclosure, preference data, trust and autonomy. But the direction is clear enough. As TechRadar has reported, AI-referred retail traffic is already concentrating around marketplaces, where structured data and reviews are abundant. That suggests merchants may be pushed towards richer feeds, stronger interoperability and more disciplined catalogue management if they want to remain visible. The commercial winner in this environment may not be the loudest brand, but the one whose products are easiest for both humans and agents to verify, compare and choose.