Hey @remotehorst23
The main work of the AI is to "process" all information of a post on the blockchain.
The reputation of the hive is deliberately ignored.
Processing is done;

• What is the syntax?
• How are the posts structured? Did the author make any effort?
• Are there the same or similar posts on the account?
• Are the words known and common or rather unknown and unique?

A normal bot will fail, because the calculated intelligence is almost 0.
If it comes to this, the AI goes through all accounts that are related to this account.
With all information it is possible to calculate at the end whether the next mail will be spam or not.

Unfortunately I can't go very deep into the topic, because this knowledge also benefits the "evil" spammers and that's not the point of this thing.
Also a lot of information are trade secrets of the CommentCoin and will be partially released later. So some more patience ;)

So I could see with a database comparison that with 64 evaluated posts you have a 54% probability that your next post will not be spam. So you would have received a CC on your next post.
Of course you can trick this now, if you know it.
But I can also see that your posts would have gotten a 55%, but your relations to bots (@reverse-engineer, @dein-problem) downgrade you a bit.

So it's a kind of own reputation through a computer that sees what's happening in real time.

There's too much entanglement.
It's not an accurate way to describe "just the Al".