Asking AI where to buy has become a habit. ChatGPT has more than 900 million weekly users, and the Gemini app has 950 million monthly users. A growing share of product research and comparison now happens in these assistants rather than on search engines or retailers’ websites.
For retailers, that changes the goal. Being named in an answer is a first step, but it is not enough. Shoppers buy from the retailer whose products the assistant can best describe and match to their needs: today’s price, what is in stock, and which model or size fits what they asked for.
Shoppers who arrive from AI are worth more
Adobe, which tracks more than a trillion visits to US retail websites, found that AI-referred traffic was up 62% year on year in July. These visitors convert 60% better than other traffic and generate 53% more revenue per visit. That makes sense: a shopper who has already researched and compared products with an assistant arrives knowing what they want, how a product could meet their needs, and is closer to buying than someone just starting a search.
A mention is not a sale
Generative engine optimisation (GEO) helps your brand appear in AI answers, and we believe it is essential. But it relies on what an assistant can read about you on the web. It can get you named; it cannot tell a shopper today’s price, whether a product is in stock or which promotion applies. GEO and access to your live catalogue work together: GEO gets you into the conversation, while the catalogue gives the assistant something useful to say once you are there.
Even the web content an assistant can access is often incomplete. Adobe’s AI Content Visibility Checker measures how much of a page an AI can read. In Adobe’s April analysis, retail product pages scored 66%, compared with 75% for homepages. That means roughly a third of a product page’s content cannot be read by an assistant. Many pages are also incomplete: information missing for human visitors is missing for an agent too, leaving it unable to tell whether a product fits the shopper’s request. Product pages are where shoppers’ questions are answered, yet they have the lowest score.
Give agents direct access to your catalogue
Retailers have several ways to become ready for AI agents, from improving product data to adopting WebMCP. Another is a retail app built on the Model Context Protocol (MCP), an open standard that lets AI assistants connect to a company’s systems. With it, ChatGPT or Claude can search your catalogue using live prices and variants, check availability by store and warehouse, build a basket and track an order. Because MCP is open, the same app works in ChatGPT, Claude and other assistants that support it.
This also gives the brand control over the conversation. Instead of an assistant paraphrasing incomplete information it found about you online, shoppers get an experience you designed inside ChatGPT or Claude: your products, your prices and the information you choose to make available.
Take a shopper looking for a smartphone at around €450. Without a catalogue connection, the assistant answers using web pages that may be weeks old. With an MCP app, it can show the models you have in stock, with your product images and logo, at today’s price and with availability near the shopper.
