How to Make Your WooCommerce Store Visible to ChatGPT, Claude, and AI Shopping Agents
AI crawlers and agents don't browse your site the way a person does. Here's what actually makes a WooCommerce store readable to them — and the specific gaps that quietly keep stores invisible.
If you check your server logs, there's a good chance GPTBot, ClaudeBot, and PerplexityBot are already crawling your store — often more frequently than you'd expect. The question isn't whether AI is looking at your site. It's whether it can actually understand what it finds.
Where AI crawlers get stuck
- JavaScript-rendered product data with no server-rendered fallback — many agent crawlers don't execute client-side JS the way a browser does.
- Price and stock only shown as rendered HTML text, with no structured markup an agent can parse reliably.
- Product images with no real URLs an agent can verify — leading some agents to hallucinate stock photography instead of showing your actual product.
- No discoverable machine-readable entry point at all — nothing telling an agent "here's a structured feed of everything this store sells."
The llms.txt convention
A growing convention, modeled on robots.txt, is a plain-text llms.txt file at your site root that gives AI agents a structured, low-noise map of what your site contains and where to find machine-readable resources — a feed, an API, an MCP endpoint. It won't replace good structured data, but it removes the guesswork about where to look.
A structured product feed
The highest-leverage single change is publishing a structured JSON feed of your catalog — price, stock, dimensions, material, color, real image URLs, and a permalink per product — at a stable, discoverable path. Suita's WooCommerce plugin publishes this automatically at /ai/products the moment it's activated, built directly from your live WooCommerce product data.
Going further: MCP
A feed is read-only and works well for crawlers building an index. For agents that need to answer a specific shopper's question in real time — "is this in stock in blue, and what's shipping to Texas cost" — a feed alone isn't enough; the agent needs to call something live. That's what the Model Context Protocol (MCP) is for, and it's worth understanding on its own.