Just like SEO ruined search, I expect companies to be running these deep researches, looking carefully at the sources, and ensuring they're poisoned. Hopefully with enough cross-referencing and intelligence models will be relatively immune to this and be able to judge the quality of sources, but they will certainly be targeted.

Or the LLM companies will offer "poison as a service", probably a viable business model - hopefully mitigated by open source, local inference, and competing models.