Here's Why Shopping Agents Might Have a Difficult Time Pulling Brands' Products for Consumers

Yahoo

13-06-2025

Agentic AI and AI-based shopping assistants continue to capture brands, retailers and consumers' attention.
But many brands and retailers may not be prepared for such a shift. While AI-based shopping assistants, like ChatGPT's new shopping function, are set up to crawl brands and retailers' sites, AJ Ghergich, vice president of consulting services at Botify, said the format of standard e-commerce sites aren't easy for agents and AI systems to ingest information from.
More from Sourcing Journal
Amazon Reportedly Tests Humanoid Robots for Parcel Delivery
Levi's Marks Three-Year Streak of Strong E-Commerce Growth
Macy's, Dick's Sporting Good Partner Grabs $44M Series B For Worker Safety Tech
That's because many sites—including those built on Shopify—display product information to consumers via JavaScript, a programming language used frequently in e-commerce to load dynamic product description pages (PDPs).
But AI agents and backend systems struggle to pull real-time information from websites running exclusively JavaScript. While JavaScript can help enhance the e-commerce experience for consumers searching directly on a brand or retailer's site, it might pose a new problem for brands and retailers as some consumers begin their shopping journeys using public AI systems, like ChatGPT. Many agents can't 'see' JavaScript in the way that humans can.
Ghergich said to help brands understand that, Botify has been showing clients how little AI can actually see.
'One of the first things we're doing [with clients] is saying, 'OK, let's look at your site with JavaScript turned off. That's what the AI is seeing,'' he said.
AI systems can crawl some generic information from JavaScript, but because the data is most often unstructured, it's difficult for them to contextualize anything about the product—the price, whether it's in stock and other important considerations. Without that information, the system is less likely to present a brand or retailer's product to a consumer, because it's unable to determine whether it fits the consumer's query parameters.
Ghergich said the technology powering the shopping assistants isn't yet strong enough to parse through unstructured data with ease.
'These bots are akin to the early days of search bots, and they can't parse this dynamic nature of modern sites yet. They probably will be able to in the near future, but today, they're blind to it,' he said.
To ensure products are included in results generated by chatbots, Ghergich and Botify suggest that brands use structured data by enabling a JSON or XML format. Typically, these formats give bots crawling the web a better chance at understanding the data, particularly when paired with a schema, which helps define subsections of the data. In using a schema, a brand or retailer can tell a bot crawling that when it uses 'price' as a label, that's indicative of how much the item costs, for example. So, a schema is how the data is labeled, and JSON or XML are how the data is stored.
Ghergich said combining these approaches won't be too tedious for retailers and brands.
'The cool thing about structured data is, once you set it up, it's literally a schema. It can go across all of your products at once, so it's one of the ultimate quick wins in technical SEO,' he said. 'You set up the schema, and now you've done 50,000 of an [action]. It goes across all of your products at once, and it's not something that you have to go page by page and manually do; it's dynamic.'
Ghergich said the next step will be better adding product details that address consumer intent—rather than simply attributes. That's because product search queries—particularly via large language models (LLMs)—continue to become longer, providing more details about why a consumer wants a specific item, rather than keywords about what they want.
For example, if a retailer had previously described an item as 'midi floral dress,' they may add data into the backend that signals how a consumer might want to wear that dress—to a summer wedding, or on vacation.
Updating product description pages with better intent may seem like a daunting task for fashion and apparel purveyors with ever-changing seasonal assortments, but Ghergich noted that it doesn't have to be done all at once—just that the transition needs to start sooner, rather than later.
'Start with your best sellers. Make sure they have those FAQ modules. Make sure that you're thinking about the customer intent in a conversational tone,' he said. 'Today's shopping journey increasingly begins with AI, not you. So if you're invisible to an AI assistant, it means you're invisible to the customer, full stop.'