When a chatter sends a message containing a banned term, they are met with the following message:
"Your message wasn't posted due to conflicts with the channel's moderation settings."

If the chatter is a harassing chatter, they will most likely just find a way to get past the banned words filter.

Example, in this scenario the word 'apple' is banned:
-> Chatter: 'Man, I really love this apple!"
<- Twitch: 'Your message wasn't posted due to conflicts with the channel's moderation settings.'
-> Chatter (trying to get past the filter): 'Man, I really love this app1e!' (notice how app1e gets past the filter)

The chatter will only try to get past the filter if they are aware that their message was withheld.

I propose a new setting the streamer can use:
-> Chatter: 'Man, I really love this apple!'
<- [The message shows up on the harassing chatter's screen, but isn't sent to anyone else]

The chatter now has no idea their message has been withheld, and the odds that he tries to get past the filter is significantly lower. Maybe the moderators even get a notification that someone has used the banned term, and can moderate accordingly.
Sometimes, this could be an inconvenience because a chatter might've used a banned term in a positive context. To solve this, the streamer could apply this setting only to a subset of banned words with a prompt "Warn the chatter when their message was withheld? y/n" or something along those lines.

This feature request is focused mainly on the banned terms filter, but the same issue arises with the automod withheld-message warning.