AI shopping agents are becoming a real path to purchase, and Shopify's Q2 numbers are some of the clearest evidence yet: AI-driven traffic to Shopify stores grew 3x year over year, orders originating from AI searches grew 3x, and AI searches powered by Shopify's Catalog converted at twice the rate of searches relying on scraped product data. If you run a catalog on Shopify, the agentic channel is no longer a pilot conversation. It is showing up in your order attribution.
This post is a working answer to the question we hear most from retailers right now: Shopify handles the Catalog for me, so am I done? Short version: the Universal Catalog is a genuine head start, and it is the starting point, not the program.
Shopify has shipped a coherent agentic commerce stack: the Catalog as the core product database (a search index now covering over a billion products), the Catalog API as the access layer agents query, Agentic Storefronts as the merchant control surface, and UCP as the transaction rails. (Scot's two-part Retailgentic deep dive walks the whole stack in detail.)
For a retailer or brand, the Catalog does four valuable things automatically. It standardizes your products into one taxonomy every agent can read. It infers attributes like color, material, key features, and technical specs. It clusters related listings, including across stores, under a Universal Product ID. And it syndicates the result to ChatGPT, Copilot, Google AI Mode, and Gemini in real time, with no per-channel feeds to maintain.
That is real infrastructure, and you should be in it. Eligibility, baseline listing, and distribution to the major agentic surfaces come essentially for free.
Look at that list from the retailer's seat, though, and a pattern emerges: everything the Catalog does, it does to your data, on your behalf, without you in the loop.
The inferred attributes are generated by Shopify's models and cannot be reviewed, corrected, or turned off. The clustering is automatic, and when it groups your SKU with the wrong variants, or another seller's listing, the error sits upstream of everything else you do.
There is a structural reason for this design. Using Shopify's reported Q2 results and Retailgentic's U.S. market modeling, Scot Wingo estimates roughly 1.1 million U.S. Shopify stores and about $265 billion in 2026 U.S. GMV. In that model, the top 15,000 stores generate roughly 80% of GMV. This is an extreme head-and-tail distribution, not a market with a representative middle. A platform operating at that scale has strong reasons to make standardized, automated defaults work for the long tail of merchants. For larger or more complex retailers in the high-GMV head, those defaults may be a useful baseline, but they may not provide the control or product-data fidelity required for differentiated agentic discovery.
And because every store runs through the same pipe, complying buys you parity: shoulder to shoulder with every competitor, listed the same way, enriched by the same models. Shopify itself is clear that inclusion does not guarantee placement or ranking. Parity is worth having. Parity is not an edge.
The gap matters most exactly where the money is. A standardized listing may cover a broad query like "hiking pants." The query agentic commerce was built for ("waterproof zip-off hiking pants for cold, wet weather in Iceland") carries four constraints, and every one has to exist, correctly, in the version of your product the agent receives. Miss or misstate a material constraint and you may be omitted, misrepresented, or less likely to be selected. Silent exclusion can happen one layer upstream of your own product page. Multi-constraint, intent-rich prompts are the whole reason shoppers use an assistant instead of a search box, and they are precisely where automatically inferred listings may not be sufficient on their own.
So the checklist for a retailer serious about this channel starts where the Catalog stops.
Control over your enrichment. Attributes, use cases, and claims you have reviewed and stand behind, not reformatted copy or unsupervised inference. This is the difference between a listing that can address a four-constraint prompt and one that cannot.
Product-level Q&A. The "is this safe for overnight use," "does it fit a carry-on" layer that intent-rich prompts draw on most heavily. In our experience, product-level Q&A is an important optimization in ACO because it addresses questions standard attributes often do not.
Your product pages, treated as agent-facing surfaces. Profound's 2026 analysis of roughly one million ChatGPT product offers found 88% were sourced from live page crawls, not feeds, and 76% were still crawl-sourced even for merchants with an integrated feed. Your product page has a new reader, and no Shopify setting reaches it.
A richer feed, ready as engines take data directly. Where an engine uses structured feed data, richer and consistent feed content can improve the quality of the representation it receives. Shopify's own 2x conversion figure shows what a stronger catalog-data path can be worth. The Catalog's feed is Shopify's version of you; you want your best version on that path too.
One product identity across both paths. Agents may pull each offer from the feed or the crawl, and you do not always pick which source is used. Consistent identifiers help ensure that whichever source the agent reads, it lands on your best representation.
Monitoring. The Catalog gives you a view into its own channel. Retailers still need to understand how agents across surfaces are displaying products, which queries they appear in, and where they may be missing or misrepresented.
This checklist is what Agentic Commerce Optimization is as an operating discipline, and it is what ReFiBuy's Commerce Intelligence Engine runs as a closed loop: evaluate, generate, enrich, sync, distribute, monitor.
We evaluate your catalog the way agents read it, across both retrieval paths, and score every SKU's readiness. We generate and enrich the full-fidelity layer the automatic pipeline cannot: reviewed attributes, use cases, and product-level Q&A, with your team in the loop instead of out of it. We sync that data into your commerce systems, including Shopify, so the Catalog itself gets a better input. We distribute your best representation to your pages for the crawl and to engines that ingest feeds directly, tied together under one consistent product identity. And we monitor how AI shopping agents actually include, represent, and recommend your products, so silent exclusion becomes something you can see and fix rather than discover in a revenue report.
None of this replaces the Universal Catalog. It compounds it. Shopify's Catalog can create a strong agentic-commerce baseline: standardized, agent-readable product data and access across the surfaces Shopify supports. ACO is the ongoing work of making that representation complete, governed, consistent across retrieval paths, and competitive when shopper intent is specific.
A practical first step: Scot built a free tool, Shop Cat Explorer, that shows the exact product data the Catalog API hands to agents for your store, brand, or product, inferred fields, clustering, and all. Check your top SKUs against your best-converting queries. What you find is the gap this post is about.
Start with the Catalog. Just don't stop there.
Shopify's Q2 2026 corporate results are reported by Shopify and discussed in Retailgentic's two-part series. Retailgentic's U.S. merchant-count, U.S. GMV, and merchant-GMV-concentration figures are Scot Wingo's estimates/modeling, not Shopify-reported figures. Retrieval-path figures are from Profound's 2026 network-log analysis of approximately one million ChatGPT product offers. For ongoing analysis of agentic commerce, Scot covers the space weekly at Retailgentic.