Model Context Protocol
An open protocol standardising how AI applications connect to external tools and data sources, letting assistants query systems directly…
AI systems that navigate websites on a user's behalf — reading pages, filling forms, comparing options and sometimes completing transactions.
Agents interact with sites through rendered pages, structured data, feeds or purpose-built interfaces, and they are far less tolerant of ambiguity than humans: unlabelled buttons, JavaScript-only flows, aggressive bot challenges and inconsistent product data all cause failures. Google has begun documenting agents that act on user request, and several vendors ship agentic checkout and research features. Preparing means clean semantic markup, accurate structured data, stable URLs, and deciding which automated actors to permit.
An agent asked to buy a specific part abandons a retailer whose add-to-cart requires a hover-only menu, and completes at a competitor.
An open protocol standardising how AI applications connect to external tools and data sources, letting assistants query systems directly…
OpenAI's agent that fetches a specific page in real time because a user or an agentic task requested it, rather than for indexing or…
Machine-readable annotations on a page, usually Schema.org vocabulary in JSON-LD, that describe entities and make pages eligible for rich…
Using HTML elements according to their meaning — article, nav, main, table, time — rather than generic divs styled to look right.
Automated, non-human requests to a site — crawlers, scrapers, monitoring, agents — which increasingly outnumber human visits and distort…