With the rise of AI agents, the internet is changing in fundamental ways. Since its inception, the internet was built for humans. When AI agents came along, those principles stayed in place, and the agents had to work around them.
They crawl as bots, browse dozens of websites in seconds, collect the information they need, and bring it back to the user. In 2026, Cloudflare Radar data showed that bots now account for over 57.5% of web requests, surpassing human traffic. Across Toll Bit's publisher network, AI bot visits grew from 1 for every 200 human visits in early 2025 to 1 for every 31 by the end of the year.
Most internet traffic now runs on principles never designed for it, and the internet can no longer effectively support the volume of AI traffic. AI gathers information from the internet in two primary ways: model training and retrieval. During model training, information is collected ahead of time; during retrieval, AI agents crawl after the user asks a question.
Retrieval is now the larger of the two. In late 2025, retrieval bots made roughly ten page requests for every one made by a training bot. Over the last year, there has been work to better support how AI collects information.
Generative Engine Optimization (GEO), also referred to as Answer Engine Optimization (AEO), is the practice of optimizing a website so AI can easily retrieve its information and is more likely to recommend it. GEO is particularly valuable for e-commerce stores, where the shopping experience is changing drastically. More and more, people are turning to AI for product recommendations.
According to Adobe, AI-referred traffic converted at a 60% higher rate than non-AI traffic in July 2026, since these shoppers arrive already referred to the brand. Every GEO project has three parts: the technical structure, the website content, and third-party citations.
Source: Towards Data Science · Summarized by HeadlinesBriefing