AI Changes the Internet: Bots Surpass Human Traffic

Bots have surpassed humans in web traffic, and this shift comes as artificial intelligence increases the number of automated requests. Data collected from various observers indicate that a majority share of online activities is now handled by software and agents, rather than by flesh-and-blood users.

The most striking reading comes from Cloudflare, which reports that the overtaking has already occurred, attributing around 57.2% of traffic to automation, compared to 42.8% from humans. Fastly, in its Threat Insight Report, presents an almost balanced but still symbolically flipped picture: 49% bot traffic and 51% human, with 99% of the bots classified as unwanted or unverifiable.

The issue is not only quantitative. Fastly notes that 47% of requests to cached content come from bots and that 60% of incoming traffic originates from automated processes, a detail that shifts the focus to scraping and access control of content. Thales, in its Bad Bot Report 2026, raises the stakes on the security front: 53% of internet traffic is composed of bots, and AI-enabled attacks have grown 12.5 times on a yearly basis.

The picture that emerges is one of a network increasingly read, extracted, and queried by automatic systems, with AI agents playing an ever-greater role in the web's underlying structure. For those managing platforms, digital services, and content publication, the issue is no longer just about blocking malicious traffic but understanding which automation to allow through and under what conditions.

For sites, this involves tracking various types of signals, such as repeated requests at regular intervals, unnatural browsing speeds, absence of typical user movements, and access sequences that follow predictable patterns. But what is the purpose of these bots? In the case of search engine crawlers, the task is to collect content and useful signals for indexing; for AI-related automated bots, the goal may be to extract text, updates, images, or other resources for training, research, or generating responses.

Welp, that happened faster than I predicted. Thought it would be the end of 2027, then early 2027, but agentic traffic is growing so fast that bots have now passed human traffic online for the first time in the Internet's history.

-- Matthew Prince 🌥 (@eastdakota) June 3, 2026

Other bots serve to simulate traffic and interactions: they open pages, generate views, click elements, fill out forms, or test the availability of services and APIs. Scraping bots primarily target data, while those used for monitoring or testing check response times, loaded content, and page functionality, often on a large scale and at very high frequencies.

The most problematic aspect concerns bad bots, which exploit automated browsing to spam, test stolen credentials, bypass access limits, or overload sites with continuous requests. In practice, they do not visit the web as a person would; instead, they read, copy, or hammer it at speeds that disrupt costs, metrics, and the security of platforms.

That said, the increase in traffic recorded in recent years is mainly due to the second type of these bots, namely those seeking information to feed the training of artificial intelligence services.