Why the same words keep turning up
A language model predicts the next likely token, so it reaches for the word that fits everywhere. Words that fit everywhere are precisely the words that say nothing in particular: crucial, robust, landscape, delve, comprehensive.
Frequency in training data compounds it. Drafts full of these words are published, scraped, and fed back into the next round of training, so the vocabulary becomes more concentrated with each cycle. That is why a phrase like rich tapestry shows up in thousands of unrelated documents.
The result is a vocabulary that signals effort without carrying information. The fix is not to ban the words, it is to notice what they are standing in for.