There is an irony worth sitting with. Companies like OpenAI and Anthropic deploy an enormous amount of automated traffic across the web. Yet as publishers of their own dynamic content, they have a sharper financial incentive than almost anyone to avoid serving bots, because every single interaction is generated on demand and costs real compute.

Contrast that with a traditional content publisher serving static media. A cached article costs effectively nothing to hand to a bot. So for two decades, static publishers had no marginal-cost reason to distinguish human from machine, and the defenses never got built. The bill for bot traffic was paid by advertisers and measurement, not by the server.

• Static publisher: serving a bot is nearly free, so bot defense is a cost center with no obvious payback.
• LLM publisher: serving a bot burns tokens and GPU time, so bot defense is direct margin protection.

This is the first time the economics of the largest content generators point *toward* human verification rather than away from it. That alignment is fragile and worth defending, because incentives, not good intentions, are what actually build durable systems.