I suppose the next internet needs to be built around networks of trust to counteract this. Everyone will be assumed to be spammers by default. Then you add whitelists and use network algorithms to determine if user X is worth your time.
Read it in the Economist -> high trust.
Any public forum -> assume nonsense.
At the time, the idea that a bot could even do that in response to any comment was pretty far-fetched. Nowadays, the flaw is obvious: How do you test that the comment is constructive?
Of course, we could train an AI - even ChatGPT - to do that (An example prompt could be "Please look at the following comment posted on a social media site. If the comment is constructive, say 'constructive'. If the comment is not constructive, say 'not constructive'. If you aren't sure, say 'unknown'. Do not give any other responses."), but at that point you're kinda fighting fire with fire.
[edit: ...how do I report a bot reply to my comment as ~~not constructive~~ spam? I genuinely cannot find any link to do so.]