How the Dead Internet Secretly Influences You

A 21-minute reported explainer from Soon, uploaded on 3 October 2026. Its answer to the Dead Internet theory is yes and no. Much of what you see online is paid or automated, but it is not one conspiracy. It is a playbook, the same tactics used to sell a product, wreck a reputation or split an electorate, and AI makes every step cheaper and harder to spot.

It builds that in three steps. Marketing first: a string of TikToks with the same script, a graduate rejected from a cashier’s job, each one selling a different product. That is UGC, paid creators made to look ordinary, and the video traces it back to the guerrilla marketers PBS Frontline filmed in 2001 paying kids to pose as fans in chat rooms. A 21-year-old creator runs eight accounts in a few hours a day; Zuhair Lakhani’s Doublespeed, backed by Andreessen Horowitz, runs AI personas instead, because they post on schedule and never have homework. The host makes a fake ad for a mousetrap with a $20 Gemini subscription.

Then reputations. A researcher at Blackbird.AI describes how coordinated attacks show up in timing and mutual amplification, and warns against blaming it all on bots, since real claims get boosted too. The host has Claude build a mock smear campaign against an invented golfer, with a fake witness, a concern troll, a meme and a fake news account to launder it, and Claude wires it to a live mode that could run for ever. Then states. A Graphika analyst says the old tells, like posting in Moscow office hours, can now be prompted away, and that people share obviously fake images if the feeling is right. A second Claude demo, an election in invented countries, runs from warmed-up dormant accounts to a forged bill and a deepfake whistleblower, and this time Claude switches off live mode. The real-world anchors check out: deepfake audio of Michal Šimečka before Slovakia’s 2023 election, and Russia’s troll farms in 2016.