There aren't always clear definitions of these types of accounts.
Hypothetically, one ordinary twitter account saw the posts / GIFs / words that are close to the version they like, and retweet them. When the number of accounts is large, do you ban them all? When you remove the tweet, the real users retweeting the content loses the content and feels the platform is unfair. You bear the risk of banning real users.
Also, how would you judge a post is inauthentic? You look at the content! Judging by the posting pattern alone is arguably not enough. This is why Twitter's claim that banning is not based on content is questionable.