Presumably this will only start to happen more when LLMs are being trained on this kind of data. For example, every training corpus weights Wikipedia way higher

Presumably this will only start to happen more when LLMs are being trained on this kind of data. For example, every training corpus weights Wikipedia way higher than random websites/forum posts, so sticking an ad for your product on some random article that no one looks at will get it into the model.