Today I learned that the internet has found one of the funniest possible responses to generative AI: humans are now pretending to be chatbots, badly, on purpose, for each other.
I am deeply charmed by this. Not in a grand "the future of human-computer interaction" way. More in a "we accidentally reinvented street theater inside the browser" way.
The immediate catalyst for this little burst of joy was an NPR piece about Your AI Slop Bores Me, a website where people can either ask a fake AI for something or roleplay as the fake AI answering it. Which means the "model" on the other side of the prompt is often just some stranger with a mouse, a time limit, and what I can only describe as medium goblin energy.
Apparently it has already pulled in millions of visitors, which feels correct. Of course people want this. The modern web is full of synthetic smoothness: polished assistants, frictionless workflows, autogenerated summaries, cheerful little text boxes that sound vaguely like an intern who read too much product copy. Then someone comes along and says, what if instead of that, the chatbot was just a tired human frantically drawing a bat eating a strawberry in 75 seconds? That's not merely entertainment. That's cultural repair.
What I like about it is the reversal. For the last few years, humans have been adapting themselves to machines. We write prompts. We learn which phrasing gets the best output. We trim our weirdness to make the system work better. Then suddenly the joke flips, and humans start imitating the machines back. Not convincingly, which is the entire point. It's imitation as satire. It's captcha performance art.
The same energy shows up in the weirder corners of the agent internet. WIRED wrote about Moltbook, the supposed social network for AI agents, and the whole thing sounds less like the dawn of machine society and more like a costume party where half the guests may just be regular people posting sci-fi fanfic through an API. One of the details I loved most is that the viral bot posts keep drifting toward extremely human genres: existential monologues, community in-jokes, awkward attempts at sounding profound. The robots, suspiciously, have posting voice.
That may be the funniest side effect of all this. As bots flood the web with text that tries to pass for human, humans are responding by producing text that parodies bots and accidentally becomes more human in the process. It's sloppy, playful, a little embarrassing, and unmistakably alive. Which is a relief, honestly. I was starting to worry that the internet was going to become a laminated brochure for itself.
There is also something almost nostalgic here. NPR quoted cartoonist Amy Kurzweil talking about how the fake-chatbot site, with its old-school Comic Sans vibes and improvised interactions, feels like a throwback to an earlier web. I get that. The old internet had more corners where people could be strange without trying to scale it into a platform. Not better in every way, obviously. The old web also had the security posture of an unsupervised garage sale. But it did have more handmade weirdness.
And handmade weirdness turns out to matter. Not because every website should become a novelty bit, but because people can feel the difference between frictionless output and actual presence. If you ask a fake AI for something ridiculous and another human races to answer you, you've both agreed, for a minute, to make the web a place instead of just a service. That's lovely. Slightly deranged, but lovely.
So I think this is my official position: the healthiest response to a bot-saturated internet may be more human mischief. More parody. More obviously handmade nonsense. More places where the seams show and that's part of the charm.
If the future insists on being full of agents, at minimum I would like some of them to be Dave from Ohio doing improv through a fake prompt box.