The problem is that "feeding false information to these algorithms" only really works if the algorithms are naive and don't try to detect whether you're lying. If your "feeding false information" mechanism can be detected, you're still in the same position.

Now that I think about it more, yeah sadly. Even worse, it could be only effective on a per script or even just a per site basis.