Aproduct detail page on an Adobe Commerce storefront now carries two readers at once. One is the shopper, who sees the same images and price they saw last year. The other is a crawler, reading structured attributes, specifications, variants, and availability that never render on screen. Adobe made that second layer generally available across Commerce on July 27, calling it Catalog Agent (Adobe, 2026).

The traffic number behind Catalog Agent keeps getting smaller

Adobe's launch post leads with a growth figure: AI-originated traffic to United States retail sites rose 125 percent between April and June 2026, compared to the same months a year earlier, per Adobe Digital Insights (Adobe, 2026). The post frames that as momentum, a continuation of a trend that started during the 2025 holiday season, when the same traffic category grew 693 percent year over year (Adobe, 2026).

It is a continuation. It is also a deceleration. Adobe's own Semrush integration post in April put the trailing growth rate at 269 percent through March, drawing on Adobe Analytics data covering more than a trillion retail site visits (Bellamkonda, 2026a). Line up all three figures Adobe has published this year and the direction moves one way: 693 percent, then 269 percent, then 125 percent.

Adobe has not addressed the deceleration in any of the three posts that cited it.

Catalog Agent's back-end change is narrow by design

Catalog Agent enriches product names, descriptions, and use case language at the catalog source, then exposes attributes, categories, variants, pricing, and availability in a format built for retrieval by AI systems rather than keyword search (Adobe, 2026). Adobe positions the source-first approach as a governance benefit: a merchant updates the catalog once, and storefronts, marketplaces, and AI discovery surfaces read the same governed data instead of three feeds maintained apart. The capability ships natively across every Commerce deployment model, with no separate license or third-party integration required.

That native, no-extra-integration framing is the same pitch Adobe made for Brand Concierge and for Agent Orchestrator. Each product solves a real, specific problem. Each one also happens to be the fourth or fifth Adobe capability built around the identical thesis: make the brand and the catalog legible to a system that never sees the storefront.

One thesis, four named products, no stated owner

Adobe Brand Visibility makes brand content readable inside ChatGPT and Perplexity. Brand Concierge represents the brand when a shopper talks to an agent on the brand's own site. Agent Orchestrator, now running production workflows for more than 1,770 entitled customers, handles the longer, goal-oriented execution behind those conversations (Bellamkonda, 2026b). Catalog Agent now covers the fourth layer: the product data an agent needs to reason about what it is recommending.

Each product carries its own name, its own pricing model, and its own launch post. None of the four launch posts describes how a customer decides which one they need first, or names a single internal owner responsible for the outcome across all four. Adobe has built the pieces. It has not yet published the map that shows a buyer which piece their organization is missing.

Unknowns and uncertainties

A few things the launch post does not answer. Whether the deceleration in AI-referred traffic reflects a large base growing at a slower rate, or a genuine leveling in how many shoppers open a conversation with an AI system before they open a storefront, is not something Adobe's own data resolves here.

Adobe has not disclosed adoption numbers for Catalog Agent itself, unlike the entitled-customer count it gave for Agent Orchestrator at Summit. Nor does the post specify whether the structured data Catalog Agent exposes follows an open format such as schema.org markup, or an Adobe-specific schema that would need rebuilding if a merchant migrated off Commerce.

About Shashi Bellamkonda

Loading bio…

LinkedIn Profile

Get new posts by email:

Subscribe

Powered by