> This means that it doesn’t work perfectly and has too many false positives and false negatives. This is unfortunately unacceptable since I don’t want to ban legit blogs, and also want to properly remove spam blogs.

I hope more platforms take note of this. AI seems deceptively suited for spam/abuse detection. Unfortunately, I've yet to see a fully automated process that doesn't make mistakes in some cases. Nearly every case I've seen of a person's obviously erroneous removal from a platform has been the result of giving automated tools too much power.

Any content based filtering has been doomed for the last 15 years.

The only long term solutions for spam are

- Cross-site reputation systems e.g. with public keys

- Paid accounts - can be also in the form of a stake that you lose in the case of breaking the code of conduc

Spam is one of the drivers that killed web 1.0 blogs, self-hosted emails, Usenet, IRC and created wallet garden social media. There is no point to keep repeating web 1.0 mistakes.

As long as Internet stays open and posting and messages are free, spam will just increase, decreasing signal/noise ratio and experience for everyone.