I don't know which models they use but it's likely already happening.

Yesterday's SEO battles are today battles to convince LLMs to produce ad tokens. The corpus is already ridden of such content. And LLM make it even easier to produce more such spam.