Every few months someone posts a screenshot of a reply section full of identical comments, and the same three words come back: dead internet theory. The idea that most of what happens online is bots talking to bots used to be a fringe forum joke. In October 2026 it is a mainstream question, helped along by AI influencers, AI slop and a war that briefly knocked a big chunk of Russia’s automated traffic offline.
A UNILAD Tech piece published on October 15 tried to put a number on it, using Cloudflare Radar’s live map of bot versus human traffic. The answer depends a lot on where you look and who is counting. Here is what the data from Cloudflare and from bot-security firm Thales actually shows, and how close the web really is to being “dead.”
The short version
- The map is Cloudflare Radar. According to UNILAD Tech’s reading, bots made up 32.6% of Cloudflare’s worldwide traffic over the past 12 months and 36.4% over the past three.
- Thales (formerly Imperva) says bots were 53% of the traffic it saw in 2025, with “bad bots” alone at 40%.
- Cloudflare says 40% of the world’s bot traffic in 2025 came from the United States, mostly out of cloud data centers.
- AI crawling is the fastest growing slice: crawling triggered by chatbot user requests rose more than 15x in 2025, per Cloudflare.
- Cloudflare CEO Matthew Prince expects bot traffic to overtake human traffic in 2027. Measured as traffic, the “dead” line is close. Measured as people, it is not.
What the bot traffic map actually shows
The map UNILAD Tech used is Cloudflare Radar, the public dashboard Cloudflare builds from the requests passing through its network. It scores every request on how likely it is to come from a bot, and lets anyone slice the result by country, city or network. Prince has said Cloudflare’s services are used by about one-fifth of all websites, which makes Radar one of the widest views of web traffic available to the public.
According to UNILAD Tech’s snapshot, Radar put worldwide bot traffic at 32.6% over the past 12 months, against 67.4% human. Narrowing the window to the past three months, bots rose to 36.4% and humans fell to 63.6%. That is a real upward drift, but roughly two of every three requests still come from people.
Zoom into individual places and the picture gets stranger. UNILAD Tech reports that over the past three months, 67.1% of the traffic Cloudflare classified as coming from Moscow was automated, with only 32.9% human. Over the past seven days, Moscow’s bot share dropped to 57%. On the other side of the ledger, the outlet says 43.4% of all bot traffic in the last three months originated in the United States.
That US figure lines up with Cloudflare’s own annual numbers. In its 2025 Radar Year in Review, the company said 40% of global bot traffic came from the US, up more than five percentage points from 2024, far ahead of second place Germany at 6.5%. The top 10 countries together accounted for 71% of all bot traffic, and Amazon Web Services and Google Cloud alone originated a quarter of it.
That matters for reading any “bots by country” map. A bot’s country is simply where its request comes from, and bots live in data centers. A country full of cloud servers, search engines or AI companies will look robotic on the map regardless of its human users.
Why Moscow looks so robotic
Moscow is a good example. It is home to Yandex, Russia’s largest search engine and cloud provider, and search engines are some of the busiest bots on the internet. When Ukrainian drones hit three Yandex data centers between October 8 and 11, automated traffic from Yandex’s network on Cloudflare Radar collapsed almost overnight. Many people on X took that as proof that Russian troll farms had gone dark.
The data does not support that leap. Most of what Cloudflare counts as automated traffic from a network like Yandex is crawlers, monitoring tools and app back ends, not sock puppet accounts arguing on social media. UNILAD Tech’s own numbers point the same way: Moscow’s bot share slipped from 67.1% to 57%, which the outlet called “hardly the dramatic fall some are painting it as.”
Thales says bots already passed humans
Cloudflare is not the only company counting. Thales, which owns Imperva and its long-running Bad Bot Report, published its 2026 Bad Bot Report in April. Covering 2025, it says automated traffic made up 53% of internet traffic, with humans at 47%. Bad bots, meaning scrapers, credential stuffers, scalpers and other malicious automation, accounted for 40% of all traffic on their own.
The trend has been climbing for years: bots were 49.6% of traffic in 2023 and 51% in 2024, when the report said they outnumbered humans for the first time in a decade. Thales also says it blocked 17.2 trillion bot requests in 2025 and that AI-enabled bot attacks grew 12.5x, from 2 million to 25 million blocked per day.
Why the numbers disagree so much
So which is it: one third bots, or more than half? Both can be true. Thales reports what it sees across the sites and APIs that use its protection products, and companies buy bot protection precisely because they attract bots: banks, retailers, ticket sellers, airlines. Cloudflare’s Radar dashboard covers a much broader and more ordinary mix of websites and counts requests for every content type, including images, scripts and API calls triggered by human visitors. Cloudflare itself notes that bot shares look different when you only count requests for HTML pages. In that narrower view, non-AI bots started 2025 with about half of all HTML requests, seven points above humans, and by December 2 humans led with 47% to 44%.
AI crawlers are the new growth engine
For most of the web’s history, the biggest bots were search engines. That is still true, but AI is catching up fast. Cloudflare’s 2025 review found that Googlebot, which crawls for both search and AI training, was responsible for more than a quarter of all verified bot traffic, and it generated 4.5% of all HTML requests on its own. Every other AI bot combined accounted for 4.2%. OpenAI’s GPTBot was next at about 7.5% of verified bot traffic.
The fastest growth is in what Cloudflare calls “user action” crawling: bots that visit a page because someone asked a chatbot a question. That category grew more than 15x in 2025. It is also where Prince sees the tipping point. “If a human were doing a task, let’s say you were shopping for a digital camera, and you might go to five websites. Your agent or the bot that’s doing that will often go to 1,000 times the number of sites that an actual human would visit,” he said at SXSW in March, according to TechCrunch. Before generative AI, he said, the internet was about 20% bot traffic. “We suspect that, in 2027, the amount of bot traffic online will exceed the amount of human traffic that’s online.”
The flip side is that AI crawlers take far more than they give back. Cloudflare’s crawl-to-refer ratio compares how often a platform crawls sites with how often it sends people to them. Anthropic’s ratio ran between roughly 25,000 to 1 and 100,000 to 1 for most of 2025 after an erratic start, OpenAI’s peaked at about 3,700 to 1 in March, and Perplexity mostly stayed below 400 to 1. Google’s search ratio, by comparison, topped out around 30 to 1. That gap is a big reason so much of the web now feels written for machines: more of the audience is machines.
| Measurement | Figure | Source and period |
|---|---|---|
| Bots, share of all traffic | 53% | Thales, full year 2025 |
| Bad bots, share of all traffic | 40% | Thales, full year 2025 |
| Bots on Cloudflare, worldwide | 32.6% (12 months), 36.4% (3 months) | Cloudflare Radar, via UNILAD Tech |
| US share of global bot traffic | 40% | Cloudflare Year in Review, 2025 |
| Germany share of global bot traffic | 6.5% | Cloudflare Year in Review, 2025 |
| Growth in AI “user action” crawling | More than 15x | Cloudflare Year in Review, 2025 |
| Growth in AI-enabled bot attacks | 12.5x | Thales, 2025 vs 2024 |
| Bot share in Imperva’s 2016 study | 51.8% | Imperva Incapsula, 2016 |
Where “dead internet theory” came from
The theory has two halves. The first is an observation: much of the activity and content online is generated by bots and pushed around by algorithms rather than by people. The second is the conspiracy: that this is a deliberate, coordinated effort, often blamed on governments, to manipulate the public. The data in this article speaks only to the first half.
The idea is thought to have started on the imageboard Wizardchan, and it spread in January 2021 when a user called “IlluminatiPirate” posted a long thread titled “Dead Internet Theory: Most of the Internet is Fake” on the Agora Road’s Macintosh Cafe forum. It went mainstream later that year with a story in The Atlantic by Kaitlyn Tiffany, headlined “Maybe You Missed It, but the Internet ‘Died’ Five Years Ago,” whose headline pointed back to roughly 2016 as the moment things supposedly changed.
That date has a statistical anchor. Imperva’s Bot Traffic Report for 2016 examined more than 16.7 billion visits to 100,000 randomly selected domains and found bots generated 51.8% of traffic. Imperva noted bots had also topped 50% in 2012 before dipping below it in 2015, so “more bots than humans” has flickered on and off for over a decade.
What changed after 2022 is the content. Large language models made it cheap to produce text, images and video at scale, and the theory’s believers suddenly had plenty of examples, from AI-generated Facebook pages to an AI influencer who built a whole fictional family on TikTok in weeks. Even OpenAI CEO Sam Altman weighed in on X in September 2025, writing that he “never took the dead internet theory that seriously” but that “it seems like there are really a lot of LLM-run twitter accounts now.”
So how close are we?
It depends on which internet you mean. Measured as raw traffic on the sites Thales protects, bots crossed the halfway mark in 2024 and kept climbing in 2025. Measured across Cloudflare’s much broader network, they sit at roughly a third and are rising. If Prince’s forecast holds, Cloudflare’s numbers will cross 50% sometime next year, driven not by troll farms but by AI agents doing shopping, research and booking on our behalf.
Measured as people, the internet is very much alive. Humans still generate most of the requests on Cloudflare’s network, and a drop in Russian crawler traffic tells you about data centers, not about who is posting on X. The more useful question for 2027 may not be whether bots outnumber us, but whether the sites we visit can still tell the difference, and whether the AI systems crawling them will send any humans back.
Sources and further reading
- UNILAD Tech: Map of bot traffic reveals exactly how far away we are from ‘dead internet theory’
- Cloudflare Radar: Traffic trends and bot versus human share
- Cloudflare Blog: The 2025 Cloudflare Radar Year in Review
- Thales: 2026 Bad Bot Report, Bad Bots in the Agentic Age
- Imperva: 2025 Bad Bot Report, How AI is Supercharging the Bot Threat
- Thales: Bots Now Make Up Nearly Half of All Internet Traffic Globally (2024 report)
- Imperva: Bot Traffic Report 2016
- TechCrunch: Online bot traffic will exceed human traffic by 2027, Cloudflare CEO says
- The Atlantic: Maybe You Missed It, but the Internet ‘Died’ Five Years Ago
- Wikipedia: Dead Internet theory
About this article: GeekBlog covers U.S. technology news, AI, phones, smartwatches and gaming. Every story is written and checked under our Editorial Policy. Spotted a mistake or have a story tip? Contact our editors.
