Dead Internet Theory Gains Traction as AI Content Surges Online

A lot of the web nonetheless runs on human site visitors, however an increasing number of it’s starting to really feel much less human.

As AI-generated posts, bots, and automatic brokers unfold throughout main platforms, researchers say the net world is beginning to resemble the issues raised by the Lifeless Web Principle, the concept that a lot of what individuals see on-line is now not produced by people however by automated machines constructed to mimic them.

When the notion first circulated a number of years in the past on conspiracy boards like 4Chan and Agora Highway’s Macintosh Café, it sounded implausible—however the rise of generative AI has modified how researchers view the declare.

In reality, bot exercise overtook human site visitors for the primary time final yr. In response to Imperva’s 2025 Unhealthy Bot Report, a world research of automated site visitors on the web, automated programs accounted for 51% of all net site visitors in 2024. AI-generated articles additionally surpassed human-written work for the primary time in late 2024, in accordance with analytics agency Graphite.



“There’s no direct technique to measure it, however a number of indicators level to the web trying totally different than we predict,” Alex Turvy, a sociologist who research how individuals work together on social media, informed Decrypt.

When researchers say bots are reshaping the web, they imply rising non-human site visitors throughout the community and the rising presence of automated or AI-generated content material inside these platforms.

The broader concern, Turvy mentioned, isn’t that fewer individuals are on-line, however that automated exercise is eroding the fundamental cues individuals use to inform who’s actual. When machines can mimic these alerts, he mentioned, customers start to doubt everybody. Some withdraw. Others transfer conversations into semi-private or gated areas.

“Lots of people are retreating to locations like Discord or non-public group chats the place they are often extra sure about who they’re speaking to,” he mentioned. “When the same old cues cease working, individuals search for different methods to know who they’re speaking to.”

That drift into non-public channels makes the general public web really feel quieter, although general human exercise hasn’t modified.

A February 2025 paper within the Asian Journal of Analysis in Laptop Science described social platforms as “machine-driven ecosystems,” arguing that bots generate 40% to 60% of net site visitors.

“These automated programs interact in scraping, spam, and manipulation, creating synthetic interactions that mimic real human exercise,” the researchers wrote. “Bots are additionally regularly employed to inflate metrics, similar to likes, shares, and feedback, fostering the phantasm of vibrant on-line engagement.”

In response to Turvy, the shift in momentum has develop into arduous to disregard.

“There’s a sign that is extra practical than we thought,” Turvy mentioned. “That’s as a result of we’re seeing the tech catch up, however we’re additionally seeing monetary incentives align.”

A September 2025 report from enterprise capital agency Galaxy Interactive discovered that automated exercise now dominates main social platforms. Analysts say the surge in AI-generated materials helps the development, noting that Reddit, YouTube, and X have seen rising ranges of repetitive, low-quality, or spam-like content material attributed to automation.

Even after Elon Musk mentioned he’d do one thing in regards to the giant variety of bots on X, by one estimate, as many as 64% of X accounts might be bots liable for 76% of peak site visitors. In the meantime, the identical research estimated that as many as 95 million Instagram accounts—9.5% of the whole—might be pretend or automated.

Meta didn’t instantly reply to a request for remark, and X doesn’t reply to media inquiries.

As artificial posts proceed to extend, researchers who observe the development say the shift is already seen within the numbers.

“About half of the web is AI-written,” Deedy Das, a accomplice at enterprise capital agency Menlo Ventures, who research the development, informed Decrypt.

“Chatbots and AI instruments summarize that materials and hand it again to you,” he added. “You find yourself studying machines summarizing different machines.”

Whereas Turvy believes the expansion of bots on social platforms will result in an exodus into smaller, extra intimate areas, Das isn’t certain the perfect of the early net can return.

“There are only a few individuals writing blogs anymore,” he mentioned. “You possibly can’t get found, and when you do, individuals assume it’s AI. Many of the dialog now occurs inside platforms constructed for efficiency, not honesty.”

Regardless of that, Das mentioned a much bigger situation is software program appearing like people.

“The web’s plumbing assumed the particular person on the opposite finish was human,” he mentioned. “CAPTCHAs, logins, two-factor codes, all of it. Now software program can imitate that completely, and there’s no shared rule for what counts as an agent.”

The rise of AI brokers

If the web feels “useless” as we speak, the unfold of AI brokers will solely speed up the development. AI brokers are autonomous packages that reply to prompts and perform duties throughout the net on behalf of a person. They browse websites, run searches, make purchases, commerce crypto, and work together with platforms in ways in which seem like human exercise.

Nirav Murthy, co-founder of mental property blockchain developer Camp Community, mentioned the sample is pushed by economics as a lot as expertise.

“Agentic AI can remix materials at machine velocity and virtually no price,” he mentioned. “Then that output goes again into circulation. Accounts begin trying totally different however appearing the identical. Engagement goes up, selection drops, and when you add human checks, the numbers disintegrate.”

As customers flip over extra management to brokers, a bigger share of on a regular basis on-line exercise will probably be dealt with by machines as a substitute of individuals, which deepens the automated surroundings they encounter on-line.

“On-line ecosystems observe incentives,” he mentioned. “When pretend engagement is reasonable and rewarded, you don’t simply get extra bots. You get manufacturing strains of automated content material chasing clicks.”

That pressure is already seen in real-world use. Earlier this month, Amazon despatched a cease-and-desist to Perplexity after discovering that its Comet browser was making purchases on Amazon’s web site by disguising automated brokers as human customers. Anthropic just lately mentioned it had blocked what it described as the primary AI-driven cyberattack, after Chinese language state-backed hackers used its Claude Code agent in makes an attempt to breach 30 firms.

The most important danger to firms and platforms, Das mentioned, comes when AI brokers are deployed in giant numbers.

“When firms run fleets of those programs, you get hundreds of thousands of requests that behave like customers,” he mentioned. “That’s tougher to see and tougher to cease.”

AI-generated video is the subsequent wave. Instruments like Sora 2 from OpenAI and Google’s Veo 3 can produce practical clips and deepfakes from textual content prompts, including to the quantity of polished however artificial content material circulating on social platforms.

Each Murthy and Turvy agreed that monetary incentives drive the flood of AI bots on-line, and proving personhood could develop into one other problem for AI to fulfill. “Humanness has develop into simply one other sign to pretend as a way to earn a living,” Turvy mentioned. “What’s lacking now’s the mess that used to show somebody was actual.”

A rising variety of blockchain tasks, together with World (previously Worldcoin), Proof of Personhood, and Human (previously Gitcoin) Passport, are rolling out programs meant to show personhood by tying on-line exercise to a verified human.

“For those who reward actual creators and make fraud costly, individuals will nonetheless have a spot on-line,” Murthy mentioned.