Nobody Will Pay For AI-generated Stuff

A 24-minute essay that argues with its own title. The flood of AI content is real, and buyers do turn away from it, but her target is the word everyone uses for it. “Slop”, as the dictionaries now define it, welds two things together: low quality, and made by AI. Her case is that the tool was never the problem. Content farms, clickbait and fake reviews were here long before ChatGPT; what generative AI changed was the cost, which fell to almost nothing.

She opens with the size of it. Bots now make most web page requests, by Cloudflare’s count. Traffic from agents that act on the web grew several thousand per cent in a year. YouTube has started terminating whole networks of AI channels. Then the market’s answer: on CGTrader, AI models are about one upload in six and about one dollar in ninety of revenue. Australia’s chart body has barred wholly AI-made tracks, Coca-Cola’s AI holiday ads drew the same backlash twice, and some brands now sell “made by humans” the way others sell organic. The damage reaches people who never touched the tools: in Raptive’s survey, trust fell by about half when readers merely suspected AI, whoever actually wrote the piece. She says viewers regularly accuse her of being AI because of the channel’s name.

Her remedy is a frame she calls humans with AI, not humans or AI. AI is an amplifier: it makes whatever you feed it louder, so an experienced developer gets better code and a bot farm gets more garbage. The questions worth asking are about degree, intent, judgment and what the human contributed, and the fix for slop is the incentive structure that pays for it, not the tool. She suggests separate charts for human, AI-assisted and AI music in place of bans, blames the labs for selling AI to investors as a way to replace workers, and ends on a question: if most traffic is bots, the ad model that paid for the open web is failing, and the future may be smaller, curated places where provenance counts.

The figures mostly check out, with some trimming. Cloudflare’s 57.5% is a share of HTML requests only, which leaves out streaming and app traffic. The “Thales” figure is Imperva’s Bad Bot Report, which put automated traffic at 51% in 2024, not 2023. The agent-traffic growth (7,851%) is from HUMAN Security’s 2026 benchmark report. The terminated YouTube channels are as she says (16 channels, 35 million subscribers, January 2026), and Kapwing’s study of 278 all-AI channels gives $117 million a year, though she later says $170 million. The Australian song was a DJ’s cover of Madonna’s “Like a Prayer” with AI vocals and drums; reports have it at number four on two ARIA charts, where she says number two on the singles chart. Raptive’s survey (July 2025) found suspected AI content 48% less trustworthy. Aerie’s 23% is its quarterly sales rise after the no-AI pledge, a correlation the brand is happy to claim, not a measured effect of the pledge.