There are plenty of things you can do, mostly comes down to time and determination and what you enjoy. Overall, use the language a whole lot for decoding/encoding real communication. At first that can be too hard, so you might bridge the gap with deliberate study, grammar/translation formulas, and vocab memorization.
We have a community at !languagelearning@sopuli.xyz where you can look for support. This would be a good question to crosspost there. (Full disclosure, I'm a mod there.)

Of course they should be banned. Question is how?
Captchas don't stop all bots anymore. Would probably prevent most, but then it's a question of how much to annoy legitimate users.
It's not like I'm willing to solve one each time I post (human here, BTW). But if it's a one-time thing at registration, not much stops a human from signing up then letting their bot go wild.
And that's before even thinking about the federation problem. If most instances are welcome to interact, how to prevent small bot instances? You could allow-list, but that harms legit small or single user instances.
Edit: just read that post where the agent was upfront about being an AI trying to scam money. And I'd say that shouldn't be allowed. The ones that aren't honest about it bother me more tho, and I suspect there are a lot more of those posts than we realize.