Banned domains are normalized into a bare, lowercase hostname and evaluated against the draft using a word-boundary regex pattern that catches full domain occurrences and subdomain matching while ignoring false-positive substrings.