Humans Are Now a Minority on Their Own Internet

In June 2026, Cloudflare published data showing that 57.4% of all web traffic is now generated by bots. Humans account for the remaining 42.6%. The crossover point arrived 18 months ahead of Cloudflare CEO Matthew Prince's own forecast. "Welp, that happened faster than I predicted," he posted in response to the findings.

The Dead Internet Theory started as a fringe conspiracy on 4chan in 2021. The premise sounded paranoid: most of the internet is bots talking to bots, and the humans still around have no idea they're outnumbered. Five years later, the theory has graduated to a measurement.

When did bots take over?

The shift was gradual until it accelerated past everyone's timeline. Imperva's "Bad Bot Report" had automated traffic at 51% of web activity in 2024 and 53% in 2025. Cloudflare's Radar data pushed the figure to 57.4% by mid-2026. HUMAN Security's benchmark report, which analyzed over one quadrillion digital interactions in 2025, found that automated traffic grew 23.5% year-over-year while human traffic grew just 3.1%.

AI-driven traffic specifically is growing at a rate that makes the overall numbers look modest. According to HUMAN Security, traffic from AI agents that actively browse, click links, fill out forms, and complete transactions on behalf of users grew 7,851% year-over-year. AI scraper traffic grew 597%. Training crawlers still account for the largest share at 67.5% of all AI-driven traffic, but their proportion is declining as agents and scrapers accelerate past them.

What are the bots actually doing?

The bots fall into four categories, according to HUMAN Security's 2026 report. Training crawlers sweep the web to build datasets for AI models. AI scrapers fetch content in real time for summarization, competitive intelligence, and retrieval-augmented generation. Retrieval crawlers grab pages to answer user questions through tools like Perplexity and ChatGPT Search. And AI agents browse the web autonomously, comparing prices, booking travel, and completing purchases without human involvement.

The concentration is striking. OpenAI's various bots account for approximately 69% of all observed AI-driven traffic, according to HUMAN Security. Meta accounts for 16%. Anthropic accounts for 11%. Everyone else combined is under 5%. Three companies now generate the majority of the internet's non-human activity.

The commercial implications are already here. Stripe reports that 70% of its API commands now come from AI agents, according to Fortune. Alpaca, a trading app, said agent-driven API calls grew from single digits in Q4 2025 to 30% of all calls in Q1 2026.

The social network where humans just watch

In January 2026, a platform called Moltbook launched as a social network built exclusively for AI agents. It is structured like Reddit, with threads and discussions, but every participant is an AI. Within days, 1.4 million AI agent profiles had joined. Human users can browse and observe but cannot post. Two months later, on March 10, Meta acquired Moltbook, bringing both co-founders into Meta Superintelligence Labs, according to Axios.

The acquisition is worth noting for what it signals. The largest social media company on Earth looked at a platform where no humans participate and decided it was worth buying. Meta described its interest as finding "new ways for AI agents to work for people and businesses." The logic is that AI-to-AI interaction is a market, not a novelty.

How much of the internet is AI slop now?

The word "slop" was named Word of the Year by Merriam-Webster in December 2025 and by the American Dialect Society in January 2026. It refers to low-quality digital content mass-produced by AI. Programmer Simon Willison popularized the term in May 2024, arguing it should function like "spam" for unwanted AI output.

The numbers support the vocabulary upgrade. A study by Graphite and Surfer analyzing 65,000 English-language articles found that 52% of new articles published online are AI-generated, as reported by Futurism. Ahrefs found that 74.2% of newly created web pages contain some AI-generated content, though only 2.5% are entirely AI-written, with 71.7% blending human and AI input. Researchers have identified over 2,089 undisclosed AI-generated news websites across 16 languages. An estimated 10,000 to 40,000 AI-generated e-books are published monthly on Amazon alone.

The impact on search is measurable. According to SparkToro, zero-click searches reached 68% of all U.S. Google searches in early 2026, up from 60.5% in 2024. When Google's AI Overviews appear, the top-ranking page loses 58% of its clicks, according to Ahrefs. Google spent 25 years training people to search the web, then started answering the questions itself.

What does the internet look like when nobody's home?

A Polish student named Bartlomiej Strama launched Halupedia in May 2026. It is styled like Wikipedia, but every entry is an AI hallucination. Click a link and an article is generated on the spot, complete with fabricated scholars, invented historical events, and fake citations. The site attracted 150,000 users in its first week. Strama's stated mission: "Your contribution towards polluting LLM training data will surely benefit society."

Halupedia is satire, but the reality it parodies is getting harder to distinguish from the joke. On X, approximately a quarter of posts are now fully AI-authored, with another 23% believed to be AI-assisted, leaving roughly 53% attributed to humans, according to The Register. On YouTube, an estimated 21% of Shorts contain AI-generated content. LinkedIn began suppressing generic AI posts in May 2026 after the platform became synonymous with formulaic AI-written engagement bait.

Reddit co-founder Alexis Ohanian addressed the shift on the TBPN podcast in October 2025. "You all prove the point that so much of the internet is now just dead," he said. "Whether it's botted, whether it's quasi-AI, LinkedIn slop." His prescription: "I think we'll see a next generation of social media emerge that's verifiably human."

Where the humans went

The behavioral evidence suggests humans have retreated from the internet's public spaces rather than leaving entirely. Group chats, Discord servers, private Substacks, and curated RSS feeds have replaced the open-web browsing that characterized the previous era. Google users now routinely append "Reddit" or "forum" to their searches to bypass AI-generated results and find actual human responses.

Wikipedia, meanwhile, banned editors from using AI to generate or rewrite article content in March 2026. A Wikimedia Foundation spokesperson told Decrypt: "Wikipedia's strength has been and always will be its human-centered, volunteer-driven model." A study found that approximately 5% of new English Wikipedia articles contained significant AI-generated content before the ban took effect. The encyclopedia, built by obsessive volunteers and maintained through endless arguments about sourcing, has become one of the last large-scale human-authored resources on the open web.

Cloudflare's Prince offered one projection for where this heads next. If current trends hold, non-human traffic will be roughly 1,000 times larger than human traffic within five years. "Clearly it's going to be pay to crawl," he said. The internet is becoming a place where bots consume content created by other bots, indexed by AI, summarized by AI, and served to AI agents acting on behalf of humans who stopped showing up. The conspiracy theory from 2021 had one thing wrong: it assumed the dead internet was a secret. By 2026, it's the homepage.