Dead Internet Theory Explained, Checked Against Bot Data

What Is the Dead Internet Theory?

The dead internet theory is the claim that most online activity is now bots and artificial intelligence (AI) content rather than real people. The conspiracy part of the theory is unproven, but Thales’s Bad Bot Report shows bots made up just over half of the web traffic it observed in 2025.

AI-written articles have also caught up with human-written ones in at least one large sample. Graphite reported that AI-generated articles surpassed human-written ones in November 2024.

To avoid AI-made content and bot traffic, use the checks in How to Find the Human Internet.

Wikipedia describes the theory as the claim that the internet “consists primarily of bot activity and automated content,” shaped by algorithmic curation. The full theory adds that this displacement of people is deliberate. The first part can be checked against traffic and content data. No evidence of a coordinated plot was found for the second part.

Where the Theory Came From

The theory was laid out in a forum thread titled “Dead Internet Theory: Most of the Internet is Fake.” A user called IlluminatiPirate posted it on January 5, 2021, on the Macintosh Cafe forum at Agora Road.

The post says the web felt emptier than it did around 2007, and it points to a turning point around 2016 and 2017. Its evidence is anecdotal: recycled threads, identical news stories and other hunches.

The post blames the United States government and tech companies. It calls the situation “artificial intelligence powered gaslighting of the entire world population.”

The idea reached a wide audience in August 2021, when Kaitlyn Tiffany wrote for The Atlantic. Wikipedia gives the headline as “Maybe You Missed It, but the Internet ‘Died’ Five Years Ago,” dated August 31, 2021.

Wikipedia says the idea may have appeared earlier on imageboards such as Wizardchan, but it flags that claim as needing a better source. Commentators usually place the internet’s supposed “time of death” around 2015 to 2016. That date is the theory’s framing. No measured event marks it.

Is the Internet 51% Bots? What the Data Shows

The 51% figure is from 2024. The latest figure is 53%, for 2025. Thales’s 2026 Bad Bot Report is the 13th annual edition and covers full-year 2025.

These bot-traffic figures count requests. They do not count people or accounts. Many bots are benign, such as search crawlers.

Thales’s 2026 report also says AI-driven bot activity increased more than tenfold, and daily blocked requests rose from 2 million to 25 million in 2025. It warns that “a much larger portion of AI-driven automation remains unverified.”

Thales/Imperva sells bot-mitigation products, and its data comes from its own network.

Imperva’s data put automated traffic at 52% in 2016 and 49.6% in 2023, according to Wikipedia. So the near-half level is not new to 2024.

Graphite used an AI detector and counted an article as AI when more than 50% of its text was predicted to be AI. In its tests, the detector had a 4.2% false-positive rate and a 0.6% false-negative rate. The false-negative test used only GPT-4o.

Graphite’s page says a newer study, “AI Now Writes as Many Online Articles as Humans,” supersedes it. Ahrefs used its own detector and says it “will never be 100% accurate.” Both studies measure only new articles and pages. Neither measures what people read.

What the Theory Gets Wrong

The bot and AI-content data support only part of the theory.

• The plot claim has no support. The 2021 post blames the United States government and tech companies. No evidence of a coordinated plot was found.
• No measured event marks the “time of death” date. Commentators usually place it around 2015 to 2016, according to Wikipedia.
• No measured source found says 60%. The highest verified figure is 53%, from Thales’s 2026 report.
• Bot share measures traffic. It does not count people. Benign automation was 13% of traffic in 2025, and bad bots were 40%.
• The forum post’s evidence is anecdotal. It rests on recycled threads, identical news stories and other hunches.

Where You Can See It Happening

Several documented cases show AI-made content reaching large audiences. Each case is one example, and none measures the whole web.

In January 2025, Meta deleted AI-run character profiles on Facebook and Instagram after users found them. Connor Hayes, Meta’s vice president of product for generative AI, had said AI profiles would exist on Meta’s platforms like human accounts. He described them as having “bios and profile pictures.”

A Meta spokesperson called the problem that stopped users from blocking the profiles a “bug.”

In March 2024, NBC News reported a wave of apparently AI-generated Facebook images, such as Jesus made of shrimp. The images drew “hundreds of millions of engagements.” The figure came from an analysis of 120 pages.

Some of those pages had been stolen from their original owners, including a Georgia church and a North Carolina high school band. The original Shrimp Jesus drawing came from a human artist, Max Arrington, who made it as a joke. Later AI copies spread, according to Scripps News.

Renee DiResta and Josh Goldstein wrote in a preprint that “AI-generated content appears to be a boon for spam and scam actors.” The preprint has not been peer-reviewed.

Their study found that Facebook’s feed sometimes showed unlabeled AI images to people who did not follow the pages. Some viewers did not realize the images were AI. Some of the pages appeared to share administrators, and the researchers said the motive was mainly profit.

Scripps News reported that Meta said it removed 964 million pieces of spam and 691 million fake accounts in the fourth quarter of 2023.

Sam Altman posted on X in early September 2025 about accounts run by large language models (LLMs). He wrote: “i never took the dead internet theory that seriously but it seems like there are really a lot of LLM-run twitter accounts now.” Futurism reported the post on September 5, 2025. Altman described what he saw on X and published no study, percentage or list of accounts.

For the broader pattern of low-quality generated content, read our guide to AI slop.

How to Find the Human Internet

The bot figures describe observed web traffic, so they cannot tell you who wrote one page. Check the page itself with the steps below.

1. Read a suspect article for signs of AI writing before you share it. Our guide on how to tell if writing is AI generated covers the checks.
2. Check a suspect image before you believe or share it. Our guide on how to tell if an image is AI generated covers that.
3. Consider which search engine you use. Our guide to search engines without AI covers the options.
4. Look for art made by people. Our guide to finding non-AI art shows where to look.
5. If you run a website, decide whether to block AI crawlers. Our guide to blocking AI crawlers covers that decision.

When a traffic figure appears, ask who measured it and what it counted. Thales counted requests, Cloudflare counted HTML requests, and Graphite counted articles. Each answer changes what the number means.

For the wider case against AI, read our explainer on why AI is bad.

Sources

• Wikipedia, “Dead Internet theory”
• Agora Road forum, “Dead Internet Theory: Most of the Internet is Fake” (January 5, 2021)
• The Atlantic, Kaitlyn Tiffany (August 31, 2021)
• Thales, 2026 Bad Bot Report (PDF)
• Imperva, 2026 Bad Bot Report landing page
• Thales newsroom, 2025 Imperva Bad Bot Report
• SecurityBrief, “Bots make up 53% of web traffic, Thales report says”
• Graphite, “More Articles Are Now Created by AI Than Humans”
• Ahrefs, “What Percentage of New Content Is AI-Generated?”
• Futurism, Sam Altman dead internet theory post
• MediaPost, “Meta Quietly Deletes AI Profiles After Public Backlash”
• NBC News, “Facebook users say ‘amen’ to bizarre AI-generated images of Jesus”
• Scripps News, “A cautionary tale: How clickbait AI images like Shrimp Jesus became a warning against scammers”
• DiResta and Goldstein, arXiv 2403.12838
• Cloudflare, “Radar 2025 Year in Review”