The most important surface for your store on Google is no longer a page you designed, and if you are still pouring your attention into the storefront while neglecting the feed behind it, you are optimizing the wrong thing. In 2026 Google folded conversational shopping directly into Search through AI Mode and into the Gemini app, and it built the buying and discovery experience for those surfaces on top of the Universal Commerce Protocol. The pipe that carries your products into those answers is your Google Merchant Center account. Google has effectively promoted Merchant Center from a back-end utility that fed shopping ads into the source of truth that determines whether an AI agent can see, understand, and recommend what you sell. That is a quiet but total reordering of where the work lives.

To understand why, follow the mechanics rather than the marketing. When a shopper asks Gemini or AI Mode for the best option in a category, the assistant does not scrape and interpret your storefront the way an old search crawler did. It draws on structured product data, and for Google's surfaces that data flows through Merchant Center. The checkout capability that lets a shopper buy directly on Google's AI surfaces runs through your existing Merchant Center feed. The product discovery capability that lets an agent pull current details about your catalog is anchored there too. This means the quality of your Merchant Center data is not one input among many. It is the input. A beautiful storefront with a starved or inconsistent feed is a store the assistant struggles to represent, and struggle usually resolves as omission.

Google has been unusually direct about what it now wants in that feed, and the direction is a decisive break from the keyword habits of the past. It announced a wave of new data attributes in Merchant Center designed specifically for discovery in conversational commerce, attributes that go beyond the traditional fields to include things the old feed never carried. Answers to common product questions. Compatible accessories. Substitutes. These are not decorative. They exist because an assistant assembling a recommendation reasons about products the way a knowledgeable salesperson would, and a salesperson needs to know what a product is for, what it works with, and what a shopper might choose instead. Filling these attributes gives the assistant the raw material to place your product correctly in an answer. Leaving them empty leaves the assistant guessing, and a guessing assistant defaults to the competitor whose data spared it the guesswork.

The consistency requirement deserves its own emphasis because it fails more stores than any single missing field. Google's guidance for the agentic era stresses that your product data must agree across every surface an agent can reach. The description on your site, the description in your feed, the details on any third-party listing, all of it should tell the same story. Your return policy, shipping information, and support details should match between your feed and your site. A human shopper glancing between two pages will forgive a small discrepancy or not even notice it. An agent treats disagreement as a reliability signal, and reliability is exactly what it weighs when deciding whether to stake a recommendation on your data. When your feed says one thing and your site says another, the assistant does not adjudicate the conflict in your favor. It moves to a source it can trust, and you never learn you were dropped.

There is a genuinely new capability Google introduced that changes how a store should think about this work, which is measurement. Google added an AI performance insights tool to Merchant Center that gives merchants a view into how their brand is performing on AI surfaces, including a comparison of their share of voice against similar brands. This matters more than it might first appear. For most of the assistant era, the great frustration has been invisibility of the invisibility, the inability to know whether you were appearing in AI answers at all, because a failure to be recommended produces no traffic, no bounce, no trace in analytics. A share-of-voice view against competitors turns that black box into something you can actually manage. You can see whether you are present, whether you are gaining or losing ground, and whether the data work you are doing is moving the needle. A metric you can watch is a metric you can improve, and this is the first time Google has handed store owners one for AI shopping.

It also pays to be clear-eyed about what platform you are on, because it changes the shape of the job. Shopify co-developed UCP and wired its stores into Google's agentic surfaces with minimal configuration, so a Shopify merchant's primary responsibility on Google is, once again, the quality and completeness of the Merchant Center data. A store on a platform that was not a launch partner has to ensure its feed reaches Merchant Center cleanly in the first place before it can worry about enrichment. Either way, the destination is the same. The feed is the battleground, Merchant Center is where it is fought, and the winner is the store whose data is most complete, most consistent, and most current when the assistant comes looking.

A useful mental shift is to stop thinking of Merchant Center as a compliance chore for ads and start thinking of it as the profile an assistant reads before deciding whether to recommend you. Every field you fill is a fact the assistant can use to match you to a query. Every attribute you leave blank is a query you cannot answer. The new question-answering attributes let you pre-empt the exact things a shopper asks in natural language, which is precisely how people talk to Gemini and AI Mode. The substitute and accessory attributes let the assistant understand your product in relation to others, which is how it reasons about a shortlist. A rich Merchant Center profile does not just make you eligible. It gives the assistant reasons to prefer you, expressed in the structured vocabulary it actually reads.

The practical program follows directly and is refreshingly concrete. Make sure your feed is reaching Merchant Center completely and without errors, because a feed that does not arrive cleanly cannot be enriched. Fill the traditional attributes fully, because blanks are forfeited queries. Then adopt the new conversational attributes deliberately, writing genuine answers to the questions shoppers ask about each product, listing real compatible accessories, and naming honest substitutes, because these are the fields Google built specifically for the surfaces you want to appear on. Audit for consistency across your site, your feed, and your marketplace listings, and reconcile every difference, because inconsistency reads as untrustworthiness. And use the AI performance insights to watch your share of voice against competitors, treating it as the scoreboard for whether your effort is working, adjusting where you are losing ground.

The larger point is that Google has told you, plainly and in its own product, where visibility now comes from, and it is not the place most store owners are looking. It is not the hero image, the layout, or the on-page copy that a shrinking share of shoppers will ever load. It is the structured, complete, consistent, current data sitting in Merchant Center, expressed in the new vocabulary Google built for conversational commerce, and now measurable through a share-of-voice tool that finally lets you see whether you are in the answer. The stores that keep treating Merchant Center as an afterthought will keep being absent from the recommendations Gemini and AI Mode generate every day. The stores that treat it as their most important page, because that is what it has become, will be the ones cited when a shopper asks. On Google's AI surfaces, the feed is the storefront, and the storefront is finally something you can measure.

Google has also signalled where this is heading in a way that raises the stakes for feed quality, through the idea of a brand agent that can converse with shoppers directly on its surfaces, functioning like a virtual sales associate. The more the assistant can act as a knowledgeable representative of your brand, answering questions and guiding a shopper through options, the more it depends on having rich, accurate, structured information about your products to draw on. A sparse feed produces a sales associate with nothing to say about you. A dense one, full of genuine answers to real questions and clear relationships between products, produces a representative that can actually sell on your behalf. This reframes Merchant Center completeness from a discovery checkbox into the raw material for an agent that increasingly speaks for your brand, and it means the enrichment work you do now compounds as these conversational capabilities mature. The stores that treat their feed as a living, complete profile rather than a static export are the ones whose brand the assistant can represent well, and being well-represented by the surface a shopper is talking to is a form of visibility that goes beyond merely appearing in a list. It is being spoken for, accurately, at the moment of decision.