Last week three important changes happened in the world of LLMs – all of which point to the increasing importance of Agentic Commerce. OpenAI made two important changes to their platform, first they significantly reduced citations of open source sites for product information and switched to relying on authentic sources of product content i.e. domain authority like a brand’s website. Second, they started relying heavily on product catalogs provided by retailers and brands as sources of product truth.

A lot of this has to do with establishing consumer trust. The leading LLMs have become much more than foundation models, they are becoming consumer (Gemini, ChatGPT, Perplexity) and business platforms (Claude, MS Co-Pilot) where consumers spend a lot of time. OpenAI has smartly realized that trust is critical to keeping consumers using their platform. For shopping, trust and accuracy are critical, neither of which were true earlier as ChatGPT sourced mostly open source data from various websites to provide product information.

Claude’s announcement of Commerce support points to them finally taking Agentic Commerce seriously. What they announced though was quite different — a toolkit for brands and retailers to build website shopping assistants, using Claude to power discoverability and leverage the Claude UCP catalog. Presumably the strategy here aligns with their push to dominate the enterprise against ChatGPTs consumer domination.

For brands and retailers, there are some very important and time sensitive changes they need to make related to discoverability of their products for Agentic Commerce:

  • Given your website has the highest domain authority as far as ChatGPT is concerned, invest in making your website content agent readable, while LLMs will scrape your side automatically, it is not sufficient. LLMs will prioritize precise, structured formats like JSON-LD. Creating content on Reddit threads and social platforms is going to be less important
  • Ensure your product catalog is enriched with use case and lifestyle data i.e. descriptions of how consumers will ask for and use the product and Q&A which instructs the LLMs on which consumer queries the product is a good fit for. Use ACP, UCP, JSON-LD and MCP (either via a ChatGPT app/plugin or as a standalone authenticated brand MCP service) to ensure this enriched catalog is available to LLMs
  • Consumers don’t buy just based on a product listing – ensure your product content is detailed and provides all of the information the consumer needs before purchase. The best ways to do this is via a ChatGPT app/plugin where your brand can provide detailer curated product and lifestyle content and advice. Forcing consumers to come to your website to find out more about the product, breaks the conversational experience

Agentic Commerce is continuing to gain momentum, investing in the above will given brands a permanent and sustained competitive advantage.