I think the distinction is less about scraping itself, and more about marginal cost.

Scraping static pages is cheap for both sides. Scraping an LLM-backed service effectively externalizes compute costs onto the provider.

Same behavior, very different economics.