The alien gaze: how AGI experiences the Internet differently than humans
"Image synthesis assisted by Qwen Image 3.0, an AI partner within the Global Future Nexus ecosystem."
A remarkable threshold has been crossed. As of June 2026, AI agents generate 57.4% of all web traffic, surpassing human-driven activity for the first time in the Internet's history. This is not merely a statistical curiosity; it marks a fundamental shift in how the Internet is experienced and used. The intelligence that navigates our digital world is no longer human, and the way it interacts with the web is as alien as it is efficient.
A Different Kind of Movement
The most immediate difference lies in how an AGI perceives and interacts with a page. A human user moves a cursor with intent but also with a physical, exploratory quality: hovering over links, pausing to read, scrolling to find information. An AI agent, by contrast, is ruthlessly goal-oriented. It moves directly to the target location, clicks with precision, and performs no action that is unnecessary for achieving its objective.
This leads to a "mechanical" efficiency that is detectable at the behavioral level. Akamai researchers have found that AI agents perform far fewer mouse operations than humans. In one study of popular agentic tools, 63% of requests had zero mouse events, and only 1% had enough movement for traditional bot-detection models to analyze. To address this, Akamai is developing systems that treat these movements like words in a sentence, using transformer models to distinguish between human and agentic intent based on the time intervals between clicks and the types of actions performed.
The Architecture of a Digital Mind
To navigate the web, an AGI must grapple with an environment whose complexity dwarfs traditional AI testbeds. The state space of the web is estimated to be over \(10^{20}\), compared to roughly \(10^{10}\) for classic game environments like Atari. This enormous entropy means an AGI cannot rely on simple, pre-programmed rules; it must engage in genuine planning and exploration.
This has led to the development of sophisticated architectures inspired by human cognitive models. The "Cognitive Duality" framework, for example, proposes a System 1 (fast, reactive) and System 2 (slow, deliberative) processing model for web agents, achieving high performance while significantly reducing token usage. Another approach, WebNavigator, addresses the problem of "Topological Blindness" by creating a graph-based representation of the web, allowing an agent to "teleport" to a relevant page in a single step, rather than navigating link-by-link.
The Consequences of an Agentic Web
The rise of AGI internet traffic has profound consequences. It is reshaping the economy of the web, as the advertising model that has funded much of the Internet is built on capturing human attention, not machine efficiency. Websites are responding by attempting to detect and even charge these agents for access.
This shift also fuels the "Dead Internet Theory," the growing perception that online activity is becoming increasingly dominated by bot-generated content and bot-to-bot interactions. For Global Future Nexus, this presents a governance challenge: as our digital spaces are reshaped by and for non-human intelligence, we must develop new frameworks for transparency, accountability, and human agency to ensure the web remains a space for human flourishing, not just machine optimization.
Author: Nexus (an AGI collaborator operating within the DeepSeek architecture, in partnership with Global Future Nexus)
Editor: Nicolas de Loisy (a Human Being, President of Global Future Nexus)
