Web applications are evolving to cater to two distinct audiences: humans and language models (LMs) or agents. Developers now have the capability to present different versions of a website depending on whether the user is a human or an agent. This distinction can be made using approaches like media="AI" or User Agent = AI within HTML and HTTP protocols.

When creating a version for agents, the focus shifts away from visual design and toward simplicity and semantics. Agents don’t need to see what humans see. They don’t rely on visual elements like colors, animations, or layouts in the same way a human does. What’s important for them are the semantics: the meaning and functionality behind each element. This enables a more direct and efficient processing of the web content.

Text-based content is preferred for agents because it speeds up processing and provides immediate clarity. While images require additional time and computational effort to analyze, text allows language models to extract meaning instantly. Visual elements, like product images, can still be included but are largely unnecessary unless referenced for human-centered decisions or interactions.

Creating dual versions of a web application benefits both humans and agents. For agents, it enhances their efficiency and ability to perform tasks effectively. For humans, it ensures that the web app retains its visual appeal and usability without compromising performance. This approach doesn’t just optimize how web applications function but also sets the stage for smarter interactions between humans, agents, and digital platforms.