Skip to main content
Solved

AI Moderation Bot - Best Practices

  • April 15, 2026
  • 5 replies
  • 60 views

Matt Breneman

Hi all!

I am setting up our Community for the first time and am exploring how to best moderate as I’m the only one, for now, running and moderating the community. I have set up the keyword blocker and also copy/pasted our community guidelines to the AI Moderator. I have tested and the keyword blocker works, but the AI moderator is not catching things that I have laid out in the guidelines. 

How do you make the AI Moderator work you? Do you copy/paste your community guidelines only? Do you treat it like an AI agent and give directions? 

 

I appreciate all of the advice you can share! 

Best answer by mitchell.gordon

Hello ​@Matt Breneman and thank you ​@JeniA for the tag!

Banned words:
I have a long list of banned words/phrases if you would like them! Setting up the banned word list seems like a weird spot to start but that is what I would recommend. A quick chart with a gpt of your choice will do a great job as well. Id recommend giving it a handful of top communities and say I would need a list of banned words for community platforms xyz. 

Moderation Bot:
Once that is completed, I would setup the AI moderator. You are on the correct track for using it in a few different ways. I break it up in sections. I give it my code of conduct. I say this is the community code of conduct: “x”. 

Next, I say these are the banned words in our community: ”x”.
 
Then, I give it Do not allow posts or replies around: ”x”

 

You should be in a very good spot after this!

Moderation sidenote, I would recommend watching tags if you allow your community to create tags. Lots of bots like to create tags and promote whatever they want. If you prevent tags from being created and have a team decision which tags are actually relevant to your community, Id say thats a really good best practice.
Happy to chat if youd like!

5 replies

JeniA
Forum|alt.badge.img+1
  • Contributor ⭐️⭐️⭐️⭐️
  • April 15, 2026

@mitchell.gordon has done awesome work getting our community to a good spot, so I’ll call him out and let him share guidance. 😀


mitchell.gordon

Hello ​@Matt Breneman and thank you ​@JeniA for the tag!

Banned words:
I have a long list of banned words/phrases if you would like them! Setting up the banned word list seems like a weird spot to start but that is what I would recommend. A quick chart with a gpt of your choice will do a great job as well. Id recommend giving it a handful of top communities and say I would need a list of banned words for community platforms xyz. 

Moderation Bot:
Once that is completed, I would setup the AI moderator. You are on the correct track for using it in a few different ways. I break it up in sections. I give it my code of conduct. I say this is the community code of conduct: “x”. 

Next, I say these are the banned words in our community: ”x”.
 
Then, I give it Do not allow posts or replies around: ”x”

 

You should be in a very good spot after this!

Moderation sidenote, I would recommend watching tags if you allow your community to create tags. Lots of bots like to create tags and promote whatever they want. If you prevent tags from being created and have a team decision which tags are actually relevant to your community, Id say thats a really good best practice.
Happy to chat if youd like!


Matt Breneman
  • Author
  • Contributor ⭐️⭐️
  • April 15, 2026

@mitchell.gordon Thank you for the tips!

For the banned words: do you have those in the Keyword Blocker section AND the the Moderation Bot? We have the keyword blocker filled out so I can use that as the banned words in the moderation section.

 

 


mitchell.gordon

Im sorry ​@Matt Breneman I think my last reply got caught in filters or I didnt hit send🤦. Yes, I do put them in both. 

 

In the moderation agent section of Community Code of Conduct, this is my setup:

  • my code of conduct:
  • Banned words / phrases in our community:
  • Do not allow posts or replies around these topics:

I use the above as headers (like above screenshot) inside. I found it better defines things for the tool and results improved.

The amount of spam that gets through, knock on wood, is .00068% with this method, in my instance. I update periodically based on what spam accounts I see register in the community and what info they put in their bios.


Matt Breneman
  • Author
  • Contributor ⭐️⭐️
  • April 16, 2026

@mitchell.gordon Thank you so much for all these tips and guidance! Super helpful!!