Inside the detection engine
How BypassGuard actually catches what others miss
Most Discord filters match exact words. BypassGuard normalizes, listens for sound-alikes, and reads between the lines — literally — before a message is ever judged.
Six ways evasion gets caught
Each layer catches a different trick — together they close the gaps a single word-list can't.
Leetspeak & symbol substitution
Numbers and symbols swapped for letters are normalized before matching — "$c4m" and "scam" are the same word to the filter.
"h4ck3r" → hacker
Homoglyph detection
Catches lookalike characters from other alphabets (Cyrillic, Greek) used to impersonate or sneak past exact-text filters.
"аdmin" (Cyrillic а) → flagged
Letter-stretching (de-stretch)
Repeated characters collapse back to their base form first, so dragging a word out no longer hides it from detection.
"sssscaaaam" → scam
Phonetic matching
Flags words that sound like a banned term when read aloud, even when every letter has been swapped out.
"fayk" → fake
Acrostic detection
Reads down the first letter of each line in a multi-line message — a banned word spelled vertically gets caught too.
first letters, line by line
ASCII-art & emoji-sequence detection
Patterns built entirely out of symbols or emoji, meant to spell something out visually instead of as text, get parsed too.
symbol shapes, not just text
Never a silent auto-ban
Detection triggers an escalation, not an instant punishment. Low-confidence matches get held for review instead of acted on — the goal is catching real evasion, not flooding your server with false positives.
Warn
First offense gets a clear, visible warning — nothing punitive yet.
Timeout
Repeated violations escalate to a timed mute, scaled to the severity.
Ban request
Severe or repeated cases get queued for staff — a human confirms every ban, always.
See it protect your server
Free to add, protecting messages within seconds. Check the plans if you want the full detection stack.
Add to Discord