You could use string patterns to filter any valid domain with a full list of TLDs, maybe filter some common replacements like a “DOT” or “D0T” followed by a TLD.
Though, some younger users don’t properly use spacing, so they might write “hello.im avigant.red is my favorite color”, and “avigant.red” would be picked up as possible free robux spam since .red is a TLD.
You are Awesome! 
Some TLDs that are also common words include:
.pizza, .red, .read, .restaurant, .tennis, .weather, and many, many more.
Many of these are perfectly valid to start sentences with so you can’t just remove any website link (even .com could be an easy typo if someone’s starting a sentence with “come” and misplace their space) without some users complaining about how unnecessary it is; the original poster’s approach might work the best if she’s quick to update her system.
Edit:
Except with how it currently works now, it doesn’t matter if she updates her system. You put the module in the game, there is no automatic updating. It’s not a require, it’s just a chat module with a table of all the values. People would have to constantly replace her system everytime, and if they added any new ones those would be removed.
Never mind then, your solution (while not perfect) would probably be more practical than manually updating the module every time a new scam website pops up.
If anything, keep the blacklist and then manually remove all urls (not the entire messages, just hash out any urls it detects) with common tlds. This would keep most of the message in tact while still censoring out the urls. Then for rarer domain names (which most scammers would try and make it see somewhat official by using a more legitimate TLD) you can blacklist them.
I agree you shouldn’t remove the entire message, though. It tends to be very misleading because the players have no idea why, or what was blocked.
Players keep reporting me that scam bots were in the game again to spam messages, I think this will be useful to resolve this case, Thankyou! 
I’m going to create an AI-based predictive algorithm that will return how likely something is a scam (0-1). I will need this for my next project so I may as well open source it when it is done to help others out.
I kinda want to modify this to add some kind of AI. Just a dumb AI that detects if a similar message is being sent a lot in a short time period.
Edit: I just realized that grilme99 has the same idea as me.
thank you i will try it out in my games. i dont like thoes bots annyways.
I have also seen instances where a bot joins send its message and quickly leaves. 
Will this still work, with the script?
Yes! This is exactly the kind of message it blocks.
Wonderful script, works perfectly! Thank you this has been a growing problem for a while now. Haven’t gotten any complaints about the scammers on the games I added it to. Also appreciate how easy to modify it is.
Okay, I was just wondering since the bots are using the /me command.
This is a great way to retaliate against the bots- thank you!
Thank you! does it work with exploited-messages
tbh a solid method would be to block messages when a user has just joined
Finally someone who utilizes the ChatService api for this aswell 
Good idea. Or make players able to chat only after they move or such.
Exploited messages are normal messages aswell, except sent by manually firing the remotes with exploits. So yes, it does work. Because, afaik, the scam bots are actually using exploits to chat those messages.