Yes. Basically the advertisers are trying to model us, and so the loop is complete in that it’s “our” standards being implemented into language via this mechanism.
However, described statistically one might say it’s reducing the variance in the distribution of language. It’s smoothing things out and removing the edge cases of language.
And one of my favorite psychology professors always likes to say that discussing anything important basically is going to offend people. So if you remove the ability to offend people, you remove the ability to discuss important topics.
The image in my mind is of a smooth ball with little jagged twig things coming off it. Filtering out offensive content is like sanding the ball and removing all the protrusions. The thing is, those protrusions are valuable because they expand the diameter of the ball.
I know that imagery is really abstract but that’s what I see in my mind as the effect of filtering out “only the offensive stuff”.