What is the dead internet theory?

The idea that most online content and activity is produced by bots and AI rather than people. The strong version claims it is a deliberate plan; the weak version, that a large share of activity is automated, is supported by traffic measurements.

Is the dead internet theory true?

Partly. Bots generate around half of web traffic, and AI makes fake text cheap. But the internet is not empty and there is no evidence of a coordinated plan.

How much of the internet is bots?

Industry reports put automated traffic at roughly half of all web requests; Imperva counted about 51% in 2024. Most of it is crawling and scraping, not posting.

Can people tell AI text from human text?

Often not reliably, especially for short casual comments. You can check yourself with the Human or AI test at hourofhumans.com/test.

Is the internet dead?

Dead internet theory: what is actually true

Dead internet theory is the idea that most of what you see online is no longer made by people: bots post, bots reply, bots like, and the humans are a shrinking audience. Part of it is a conspiracy theory. Part of it is now just measurement.

Where the idea came from

The phrase spread from a 2021 forum post and grew as people noticed the same strange things: comment sections full of near-identical replies, accounts that never sleep, viral posts that seem to be written for nobody in particular. The strong version claims the internet "died" around 2016 and has been mostly fake since, steered on purpose. There is no evidence for a plan like that.

What the numbers say

The weaker version, that a large share of online activity is automated, is not a theory any more. Security companies that watch web traffic have reported for years that bots make up close to half of all requests. In 2024, by Imperva's count, automated traffic passed human traffic for the first time in a decade, at about 51%, and a large part of it was "bad" bots: scrapers, fake sign-ups, ticket and sneaker scalpers, credential stuffing.

Traffic is not the same as conversation, though. Most bots crawl and scrape; they never write a comment. The part people actually feel is different: since 2023, language models made it cheap to write text that reads like a person. A fake review, a forum reply or a "my honest experience" post now costs almost nothing to produce at scale.

What is exaggerated

The internet is not empty. Billions of people still post, message and argue every day.
There is no single group running it. Spam, ad fraud, engagement farming and propaganda all point the same way without needing a plan.
Bots are not everywhere equally. Small forums, group chats and places where people know each other are still mostly human. Large open feeds, review sites and reply sections are where automation gathers.

How to check for yourself

Take our Human or AI test: ten short comments, some written by people before 2019 (before AI could write), the rest by current AI models. Most people expect to spot the AI easily. Try it before you decide how dead the internet is.

Look at the account, not the text: age, gaps in activity, whether it ever replies to replies.
Ask something specific and local. Generic answers to specific questions are a warning sign.
Prefer places where people meet offline too. A real name is not proof; showing up is.

If text cannot tell, what can?

If reading cannot separate people from machines, the remaining test is the physical world. A bot can write a perfect comment; it cannot stand in a park on a Saturday. That is the idea behind Hour of Humans: an anonymous count of real people. You tap to get a number, and once your city has enough people, you show up for half an hour and a few strangers confirm you exist. No name, no face scan.