I think what we’re concluding here is that using Google to obscure the linguistic style is flawed, because a state actor could obtain the original linguistic style from Google records, or from their own records of snooped traffic.

In other words: the blog should find a way to obscure linguistic style offline.