The Dead Internet Theory is basically coming true

If social media feels a little hollow lately, you are not imagining it.

The "Dead Internet Theory"—once a fringe conspiracy claiming the web is populated mostly by bots mimicking humans—is rapidly becoming our day-to-day reality, according to researchers and new platform data.

Machines are outnumbering humans online

Automated systems now generate more traffic than actual people.

According to Imperva’s 2025 Bad Bot Report, automated traffic hit 51% of all web traffic in 2024, crossing the human threshold for the first time. Meanwhile, analytics firm Graphite found that AI-generated articles surpassed human-written posts late last year.

Menlo Ventures partner Deedy Das noted that roughly half of the internet is now written by AI. Because chatbots are constantly scraping and summarizing those very pages, users increasingly end up reading machines summarizing other machines.

Social networks are feeling the brunt of the shift. A September 2025 study from Galaxy Interactive found that automation dominates major feeds. Outside estimates suggest up to 64% of X accounts might be automated, generating 76% of peak traffic, while Instagram hosts around 95 million fake accounts.

Sociologist Alex Turvy told Decrypt that humans haven't logged off; rather, the cues that prove someone is real have broken down. The result is a massive retreat into gated communities like Discord and private group chats.

When agents disguise themselves as people

The bigger headache is autonomous AI agents that can browse, trade, and shop just like people do.

The entire web plumbing—from CAPTCHAs to two-factor auth—assumed a human was sitting at the keyboard. AI agents shatter that assumption.

Tensions are already boiling over. Amazon recently sent a cease-and-desist to Perplexity after discovering its Comet browser disguised automated shopping agents as human customers to make purchases. Anthropic also revealed it recently blocked what it called the first AI-driven cyberattack, where attackers leveraged Claude Code agents to hit 30 companies.

With hyper-realistic video generators like OpenAI’s Sora 2 and Google’s Veo 3 flooding feeds next, proving personhood via identity networks like World or Human Passport is suddenly a serious conversation.

Why it matters

For everyday users, the open internet is turning into high-friction synthetic sludge. Looking up unvarnished reviews, product recommendations, or genuine advice on public forums now requires wading through AI-generated filler.

In our view, creators and businesses can no longer count on public algorithms for genuine engagement. The smart move right now is building direct, authenticated relationships—via email lists, private servers, and real-world touchpoints where human identity is clear.

For platforms, this trend threatens the bedrock of digital advertising. Ad models built on views, clicks, and vanity metrics make zero sense when automated agents can generate endless engagement for pennies.

Until web standards evolve to explicitly identify and govern software agents, companies will be forced into awkward, whack-a-mole legal fights like Amazon's clash with Perplexity. The open web wasn't built for this, and it shows.