The Dead Internet and Your Pattern-Hungry Brain

You open a comment section and something’s off. Forty replies, every one weirdly agreeable, every one phrased like a press release that learned to smile. The identical five-word compliment under a hundred unrelated videos. A reply guy whose grammar is too clean for someone who just rolled out of bed. You can’t prove a thing, but a small voice asks: is anyone actually here?

That voice has a name now. Dead Internet Theory is the half-joking claim that big stretches of the web aren’t humans talking to humans anymore, but bots talking to bots while the rest of us scroll through the wreckage. As literal conspiracy it overreaches. As the description of a vibe, it’s onto something honest, and worth taking apart.

Start with the half that’s real. I’m a cybersecurity junior, so the bot part doesn’t shock me; automated accounts, content farms, and engagement rings have been load-bearing infrastructure for years. What changed is the cost. Maggie Appleton has a phrase I can’t shake: people are “jetspraying the web with AI slop.” When a thousand plausible paragraphs cost less than a coffee, the ratio of synthetic to human tips hard, and your unease is just tracking a genuine shift in supply.

Now the half that’s your own wiring, which nobody warns you about. Your brain is a pattern-detector that never clocks out, and it’s tuned to err toward false positives. Psychologists call the habit of seeing meaning in noise apophenia. It kept your ancestors breathing (mistake a stick for a snake and you lose a second; mistake a snake for a stick and you lose the rest), but online it means that the moment you expect bots, you start finding them everywhere, including in flesh-and-blood people having a bad-grammar morning. The Dead Internet feeling is real signal pushed through a brain built to over-detect.

There’s a second mechanism stacked on top, and this is the one to actually remember. The illusory truth effect, first measured by Hasher and Goldstein in 1977, says we rate a claim as truer for the simple reason that we’ve seen it before. Repetition impersonates evidence. Now picture a web where one machine can repeat a single claim ten thousand times before lunch. The slop doesn’t have to win the argument. It just has to get familiar, and familiarity strolls right past your reasoning through a side door you didn’t know was unlocked.

You can’t out-scroll this, and assuming you’re personally immune is just the trap wearing a clever hat. The only real defense is metacognition: catching the warm click of “this is obviously true” and interrogating where it came from. Did I reason my way here, or did I just see it forty times? That question is the entire game. More on the foraging tax of all this noise in ai-slop-and-the-cost-of-foraging, and on training the inner observer that asks it in metacognition-eileen-gu-and-going-public.