The Accessibility Tree: A Subtle Lever for AI Commerce

AI agents don't see your PrestaShop store the way you do. They read the accessibility tree, a structured representation built by the browser. OpenAI confirms that ChatGPT Atlas interprets ARIA roles and labels to understand interactive elements. Similarly, Microsoft's Playwright MCP relies on accessibility snapshots rather than screenshots. For a merchant, this means an "Add to cart" button implemented with a simple <div> tag and an onclick becomes a generic node without a name for the machine. The agent can neither identify it nor act on it. Replacing that button with a native <button> element with an accessible name allows the AI to recognize the action and, eventually, execute it via standards like WebMCP (currently in a Chrome origin trial). Concretely, audit your product pages using Chrome DevTools' tree inspector or an extractor like AXray: identify interactive elements without names, missing headings, and fix them in the template. The SEO effect is indirect: a well-structured site facilitates content extraction by AI agents, which can influence their ability to cite your products or guide transactions. But the real impact still depends on the adoption of these agents.