I'd tested a bunch of words and phrases on a privated video of mine before. The removal time is 20 seconds if it is forbidden on YT. "Homosexual" still works on that video, at least. So it's not quite top-tier restricted yet. But there is a lot they don't allow to be posted at all, profanity filter turned on or not.

Honestly, this "feature" has been annoying me for months. It's getting to the point where I might gave up on commenting about gender ideology related issues on YT altogether. Because that seems like the only time this latent filter kicks in. I don't really use swear words in comments, but terms and phrases on this subject are frequently restricted as if they were profanity or hate speech. Often times I write something long and forget to copy the text, I eventually find out it was removed after posting and no one saw it 😠

I then spend the next long while re-writing and testing the offending passage in the test video is determine which parts I should excise. What makes it harder is that YT remembers the passage and blocks it even when you have removed the only conceivable "offensive" word/phrase and even modified the text layout. It's some neural-net machine learning. Only Google is capable of controlling speech in this manner at this level. This is why moving away from the Google ecosystem is something I'm trying to do now.