Agentic commerce is shifting digital shopping from “search and click” journeys to AI-led decisioning, where products are recommended or purchased based on how well they match intent. While AI agents may look similar at the interface level, the key difference is the commerce environment: Google is building agent-led shopping within its own ecosystem to protect monetisation and measurement, while open agent ecosystems can route shoppers across retailers and brand sites.
For brands, this creates a new competitive battleground: discovery is now driven by micro-signals like use-case tags, compatibility, performance claims, and sentiment, not just keywords and PDP quality. The implication for PIM and PXM is clear. Structured, decision-ready product truth is now essential to be included in an agent’s candidate set, and PXM must scale enrichment across every surface where agents learn and reason.
Early evidence shows this is already commercial reality, not a future scenario: leading retailers’ AI assistants are reaching material adoption in peak periods and driving meaningful incremental sales. At the same time, platforms are building semantic commerce layers that reason about use cases, compatibility, and fulfilment reliability making the PDP more important than ever as the machine-readable backbone for discovery and trust.
For years, digital commerce optimisation meant the same thing: win search visibility, drive clicks, convert on the product page.
That model is changing quickly.
Today, AI is no longer just helping shoppers browse. It’s increasingly making decisions on their behalf, turning product discovery into a machine-driven process where relevance is calculated, not searched.
It’s no longer enough to be searchable. You have to be selectable.
Traditional ecommerce discovery was largely driven by:
Agentic commerce shifts this toward intent interpretation and decision execution.
Instead of a shopper typing a query, clicking ten links, and comparing manually, an agent can interpret the goal, break it into criteria, evaluate options, and recommend or complete the purchase.
Adobe data shows 39% of U.S. consumers have used generative AI for shopping, with AI-assisted journeys showing conversion rates up to 12.3% versus 3.1%.
Across major retailers, we’re seeing a consistent architecture emerge: visible conversational agents sit on top of invisible ‘commerce intelligence’ layers that actually understand products, use cases, and constraints. These semantic layers — such as Amazon’s internal product graph or Google’s Shopping Graph are doing the real reasoning work behind the scenes.
The agent reasoning layer is becoming similar across players; what differs is the commerce environment.
Open ecosystems: intent fulfilment across the web
In open agent ecosystems, AI can evaluate products across sources and direct shoppers to a retailer, marketplace, or brand site — or complete checkout in-chat.
Platform ecosystems: monetised decisioning inside walled gardens
Google’s Universal Commerce Protocol (UCP) standardises merchant capabilities such as payments, shipping, discounts, consent, and post-purchase flows.
This mirrors a broader industry pattern: a separation between the customer-facing agent interface and the semantic product layer that actually powers selection. Brands must therefore feed these intelligence layers with rich, machine-readable product truth, not just persuasive copy.
By keeping commerce inside its ecosystem, Google can monetise recommendations, optimise for conversion, and retain full attribution and measurement.
Discovery is the battleground
Discovery is no longer a single query problem; one request becomes many micro-queries around materials, performance, sizing, price, use case, and sentiment.
In other words, the gating factor is no longer ranking it is agent eligibility. A single missing attribute can remove a product from consideration before any comparison begins
PIM becomes decision infrastructure: complete, standardised, validated, and enriched with the dimensions agents query.
Platforms are investing in semantic product graphs that connect attributes, imagery, reviews, availability, and behaviour so agents can reason about outcomes, not just keywords.
Google’s UCP is both a technical and strategic move to remain the gateway for AI-led shopping, while open protocols route commerce across ecosystems. The next frontier will be ‘Agent Optimisation’ structuring product truth so machines can confidently understand, trust, and act.
Final thought: In agentic commerce, visibility is no longer about being present — it is about being chosen.