Yes, this trend has been becoming more observable - particularly in the past ~10 years IMO. Of course, as another commenter points out, it has been increasingly adopted by some politicians themselves.
At some point, all of the "megaphones" we handed every single person with access to the internet, started to be more and more co-opted by actors whose primary use is propaganda, trolling, etc. And, particularly, in the sense of what has been termed "coordinated inauthentic activity" and the like. So, whereas, 15+ years ago you could look at comments on YouTube, say, and see a mix of discussion and some 'low quality' comments of various types (like trolling, random nonsense and non sequiturs, etc.), more recently, there are specific efforts to hijack comment threads, "stir the pot", etc.
A lot of the "bad actor networks" hinge, as usual when it comes to humans / statistics / networks, on a small set of "high volume central nodes". Although, apparently, that has been evolving as well, presumably as a sort of war continues to play out between companies hosting various platforms, and people, groups, countries, etc. who wish to exploit their availability. [1]
There are two potential mitigations I see forming more and more. One appears to be proven in a practical setting, already, though it has negatives that have also been fairly substantially discussed. The other involves a tool gaining more and more ground only even more recently:
1) Turn off / remove general (and especially unmoderated) "comment" features*
2) Use increasingly sophisticated ML and other models to handle various types of activity and language that are being deployed for corrupt purposes.
The problem with "1" that is most frequently cited is that this can push "networks of bad actors" more into the shadows - where it's harder to know what they are up to etc. However, this is obviously a trade-off against their reach in general etc. In the shorter-term (and, I personally think, likely even in medium- to longer-term), this clearly "works". It substantially reduces harms to individuals, various social groups, societies as a whole (arguably), etc.
The second option is of less clear value, so far. Automated moderation has been around for 30+ years - that's about the time I have personal experience with. It's only in the past 7 - 10 years that more sophisticated statistical models etc. seem to be being deployed (AFAIK). And, we are now in a period of really rather more significant change with ML capabilities and the like. I'm inclined to think we're still not at the level required to make this "easy" / "silver bullet". But, I also think there are likely significant improvements being implemented and seen in practice, currently. Of a type that is relatively "transparent" to most users ... so, not being involved in any such efforts, not having spent time looking into these efforts more closely, I'm entirely speculating here.
In any case, much as "publishing" went through periods of less and more "regulation"**, we've had a reasonable period of very low "regulation" of the internet. Regulation need not, of course, mean government regulation. But, it seems increasingly likely that regulation in some form or other will increasingly need to be imposed on the internet (at least the "mass / easily accessible" parts) to diminish the influence of bad actors and keep the internet valuable for everyone (else).
* Obviously, this depends on nature of platform - for many platforms, these features are far from core. Something like Twitter/X is, of course, pretty much "pure comments" ... so, irrelevant there, regardless of who is running it.
** There were periods where pamphleteers in America (where I'm more familiar with the history) were running wild in political and other discourse, and then this was brought under more of a semblance of and practical "moderation"
[1] E.g., https://dl.acm.org/doi/abs/10.1145/3522756 (just one of a slew of relatively recent papers, more and more being published each year, of course)