When a shopper asks an AI assistant what to buy, no salesperson is in the room. The assistant decides based on what it can read about your products. Your catalog has quietly become your most important sales rep.
For two decades, ecommerce discovery worked through the search box. Shoppers typed keywords, platforms ranked results, and brands competed for position with relevance, reviews, price, and advertising. That model is not going away, but a second one is growing alongside it.
Shoppers now ask questions. “Which water purifier is best for hard water in a two-bedroom flat?” “What is a good gift for a ten-year-old who likes science?” They ask Amazon’s Rufus, Flipkart’s Flippi, ChatGPT, Gemini or Perplexity, and increasingly they get a shortlist with reasons attached. Some assistants are even starting to complete purchases.
In that moment, no brand presence exists in the conversation except your product data. Whatever the assistant can read, understand and trust about your products is your entire pitch.
What a shopping agent needs to recommend you
None of the major assistants publish exactly how they choose products, and brands should be wary of anyone claiming a formula. But the inputs they can draw on are visible, and they point to what a catalog needs.
Complete structured attributes. Agents match products to requirements. If a shopper asks for a purifier suitable for high TDS water and your listing does not state the technology, capacity, and TDS range in structured fields, you are hard to match even if your product is ideal.
Answers to real questions. Agents look for information that answers the shopper’s specific question. Use cases, suitability, compatibility, and limitations stated clearly in your content give an assistant something concrete to repeat.
Consistency across sources. Agents read your marketplace listings, your website, reviews, and third-party content. When those disagree on size, ingredients or warranty, the assistant has to choose, and it may choose the wrong version or none at all.
Evidence of quality. Ratings, review themes, and answered questions all shape how an assistant describes a product. Recurring complaints in reviews tend to appear in AI summaries too.
Why most catalogues are not ready
Product content in most brands was built for human shoppers scanning a page. Titles are packed with keywords, key facts sit inside images, bullets lean on adjectives rather than specifications, and attribute fields are filled only where a platform demands it. Each marketplace version has drifted from the others.
That worked when a person looked at the page and filled in the gaps. An agent cannot read a benefit printed on a lifestyle image, and it cannot assume that “long lasting” means 48 hours.
Preparing the catalogue
- Create one master product record. Store every fact about every SKU once and approve it, so every channel draws from the same truth.
- Fill attributes exhaustively. Treat optional fields as mandatory. Dimensions, materials, ingredients, compatibility, certifications, care instructions, and intended users all help agents match products to needs.
- Write for questions. Add plain language answers to the questions shoppers actually ask, drawn from reviews, customer Q&A and support logs.
- Move facts out of images. Anything important should exist as text. Images support the story; they should not be the only place a fact lives.
- State limitations honestly. “Not suitable for induction cooktops” is useful information for an agent and prevents a wrong recommendation that would end in a return.
- Keep channels aligned. Audit live listings regularly against the master record, because content drifts.
Measure what the agents say
Traditional search gave brands rank and share of search to track. Agentic discovery needs a new habit: ask assistants the questions your customers ask, and record what they say. Are your products recommended? Are they described accurately? Which competitors appear instead, and why?
This is the practical core of generative engine optimisation for ecommerce: understanding how AI assistants see your brand and products, then improving the content and data they draw on until the answers they give are accurate and favourable.
The sales rep you cannot brief in person
A good salesperson knows the product, answers honestly, understands who it suits and never contradicts the spec sheet. That is now the job of your catalog. You cannot brief a shopping agent before the conversation or correct it afterward. You can only make sure that everything it reads about your products is complete, consistent and true.
Brands that treat product data as a marketing afterthought will find themselves left out of more and more shortlists. Brands that treat it as their sales force will be the ones the agents recommend.
