AI Shopping Agents Need Cleaner Catalogs, Not Louder Hype
AI shopping agents are software assistants that use large language models and structured product data to search, compare, and recommend items across many retailers on a shopper’s behalf, often handling discovery, evaluation, and even checkout inside a single conversational experience.
The core story is not that AI shopping agents exist, but that they are only as smart as the catalogs they read. Retailers are rushing to make listings visible to these agents, yet many still cannot tell when two pages describe the same product. That is a structural failure, not an AI failure. Shopify’s new Catalog system moves the industry in the right direction by treating catalog management and product deduplication as first-class problems, using LLM product data pipelines instead of brittle rules. As product discovery shifts away from manual browsing towards AI-guided journeys, the brands that will win are not those shouting the loudest about agentic commerce, but those quietly fixing their data so that agents can reliably represent their assortment.
Inside Shopify’s LLM Catalog: Teaching Machines to See ‘One Product’
Shopify’s Catalog is a bet that AI shopping agents cannot do their job until the underlying product graph makes sense. It uses LLM product data pipelines to read messy merchant feeds and reorganize them into coherent, agent-ready structures. In one example, two merchants sell the same protein powder: one as a single listing with flavor variants, the other as separate listings per flavor. Humans know this is the same product; machines historically do not. Catalog groups related listings under a Universal Product Identifier so agents recognize when different pages describe the same underlying item and can compare them directly. The system goes further by inferring a product’s “core value proposition” — the main reason a shopper buys it — to decide which listings belong together. This is catalog management turned into a language problem, and that is exactly why LLMs are the right tool.
When we get this right, merchants’ products show up exactly where they should across every agent, which helps them get customers from new agentic distribution channels. That is the economic incentive: if your data is clean enough for Catalog to deduplicate, AI shopping agents can identify and compare identical products across retailers more accurately, instead of burying you behind duplicate clutter or misclassified variants.
Why Retailers Are Scrambling: Agents, Abandonment and Standards
Retailers are not panicking about deduplication for fun; they are reacting to a clear shift in shopper behavior. AI-driven traffic to Shopify-powered stores grew eight times year over year in the first quarter of 2026, while orders from AI-powered searches increased nearly 13-fold. New buyers through AI channels are arriving at nearly twice the rate of other channels. At the same time, 53 percent of consumers who used generative AI for search also used it to help them shop, and yet cart abandonment still hovers around 70 percent globally. That combination — agents driving discovery while customers remain overwhelmed — makes product deduplication a survival issue. If an AI agent cannot distinguish between your real offer and a dozen near-duplicates, it will not present you cleanly in shortlists.
This is pushing brands to rethink catalog management from the ground up. One search and AI visibility executive says “the thing I can’t get off the phone about right now is some version of an agentic catalog,” whether that means adopting Shopify’s tools or building their own with partners. Proper product data organization is no longer a back-office concern; it is the interface between retailers and AI shopping agents. That is why standards like the Universal Commerce Protocol, which standardizes how agents create carts and complete payment across merchants and has backers including major marketplaces and big-box retailers, are attracting early adoption.
Agentic Commerce Meets Fashion: Data, Loyalty and Voice
Luxury and fashion brands are learning that agentic commerce is not an abstract future, but a new layer between them and their most loyal customers. On June 16, senior commerce and fashion executives gathered in New York for a dinner marking the launch of an “agentic storefront” built to bring AI to the centre of the customer journey. The discussion was not about chasing every new protocol, but about where to invest as more of the shopping journey is mediated by AI. Leaders from houses and contemporary labels alike framed loyalty as both an emotional relationship and a data problem, investing in technology to “make sure we know who [customers] are, listen to what they want and react to it.”
These brands are also experimenting with how AI agents should sound. One agentic storefront launching soon will use a designer’s own voice as the agent’s voice, making the brand itself — not a generic assistant — the main interface. That is the crux of fashion’s response: embrace agentic commerce without handing over the relationship. The industry understands that even as AI shopping agents compress discovery into a few prompts, customers still come back for a sense of being known and a clear alignment with brand values. Clean product data is the ticket into the AI conversation; distinctive human voice is what keeps shoppers listening.
From Hype to Infrastructure: What Comes Next
The promise of agentic commerce has moved faster than its reality. Early experiments with direct checkout inside conversational assistants have already been walked back in favor of discovery-first experiences and merchant-controlled checkout flows. That should be a warning: bolting AI on top of brittle catalogs will not fix broken shopping journeys. The next phase is infrastructure, not spectacle. Botify is piloting its own agentic catalog product with several large retail brands, and many are planning for their back end to speak directly to emerging protocols like the Universal Commerce Protocol.
Retailers who treat LLM product data pipelines, product deduplication, and agent-ready catalog management as core capabilities will be ready when agentic commerce becomes mundane rather than novel. Those who keep their data fragmented will watch AI shopping agents steer customers toward cleaner, more comparable assortments. The playbook is clear: organize your catalog so machines can see one product where humans do, adopt open standards so agents can transact, and design AI experiences that sound like your brand instead of a generic bot. The agents are coming either way; the question is whether they will know what you are selling.
