Matthew Widdop
Matty is a Digital SEO Executive at Intelligency, helping our clients to improve their digital performance. Matty is currently studying for his Level 3 Multi-Channel Marketer apprenticeship and has completed a Sport Journalism Degree at the University of Huddersfield.
AI Agent Testing can connect AI to a site, giving it the ability to test AI machine readiness, including whether systems can extract and understand content and whether there are any technical obstacles.
Another function of AI testing is that it also has the ability to perform tasks on a site, such as making a purchase, filling out a form, etc. It can then report on any issues it faces in this regard; this is important as users may start using AI to complete tasks for them more regularly in the future, not just using it as a research tool.
How does it work?
Agentic testing works by connecting an AI Agent to your website using a Model Context Protocol (MCP). The best way to describe an MCP is that it’s like a USB for AI. It gives AI the ability to connect to different software, files or browsers. Chrome DevTools is a set of web development tools provided by Google Chrome. By connecting AI to Chrome DevTools MCP, the AI can then use these tools to analyse your website, report on its findings, and complete tasks such as purchasing, ordering, newsletter sign-up, etc.
What can you learn from AI Agent Testing
- Crawlability and discovery – Can the website’s content and pages be discovered by AI
- Machine readability – Can systems extract and understand the information, H1s, Titles, descriptions, and Schema
- AI Search content readiness – Editorial changes to improve chances of content being used in AI
- Technical and AI-agent obstacles – Technical problems stopping AI from completing tasks on the site
- Improved Conversion Journeys – Can AI interact with the site such as purchases, form fill-outs, etc.? Is everything working properly
How to set up AI Agent Browser Testing
- Install Node.js – Node.js lets you run javascript locally on your computer, which you will need to run the JavaScript program that powers the Dev Tool MCP.
- Install AI Agent – For some of the more popular AI agents such as ChatGPT and Claude, you can’t always connect an MCP to the web version of the app. In such cases, you will need to either download the desktop version or connect to them via your computer’s command centre, PowerShell. (If you are on Microsoft)
- Set up Chrome Dev Tools MCP – Connect Chrome Dev Tools to your chosen AI Agent. If using the desktop app there should be some built-in app connections or you may need to connect it manually using PowerShell commands.
- Prompt the AI – Once your AI Agent and MCP is fully connected, you need to make sure you prompt the AI correctly so it pulls through the sort of information you want, but don’t
Results
AI Agent Testing provides businesses with a way to understand how prepared their websites are for an increasingly AI-driven internet. By connecting an AI agent to Chrome DevTools MCP, businesses can assess how easily AI systems can extract and understand their content, identify technical issues and test whether AI agents can complete tasks such as navigating pages, filling out forms and making purchases.
