HOW THIS WORKS
Carriers run automated content filters on every A2P text. The engine applies the same kind of heuristic a carrier filter does — it reads the words in your draft and flags the combinations carriers treat as risky. It's a risk estimate, never a delivery guarantee.
The principle: context, not just keywords
A word that's harmless in one message can be risky in another. The clearest example is your own product: knife is fine on its own. But the moment a blade term appears in the same message as a firearm term — "gun", "rifle", "9mm", or a gun-show name — carrier filters read the pair as a weapons pitch and quietly drop the message. The engine calls this a co-occurrence trap, and it is the single most important thing this tool catches.
New Cutco knife set is in stock — stop by booth 12 this weekend.
"Knife" alone is your legitimate product. No trap fires.
Gun and knife show Saturday — bring cash for the best deals on ammo.
"Gun" + "knife" + "ammo" in one message trips the gun + knife co-occurrence rule.
What gets flagged
The engine sorts risky words into eight categories. Each category carries a per-hit penalty that lowers your score; co-occurrence traps stack an extra penalty on top.
Firearm & ammo
−28 pts / hitGuns, handguns, rifles, shotguns, ammo, magazines, calibers (9mm, 45 ACP), suppressors, AR-15/AK-47, concealed carry, holsters. Heavily filtered even as a passing mention.
Generic weapon terms
−24 pts / hit'Weapon', 'lethal', 'kill', 'stab', 'assault', 'tactical', 'combat', 'militia', 'self-defense' used as a weapon pitch.
Blade / edged (only with a trap)
−10 pts / hitKnife, blade, dagger, machete, sword, switchblade. Low-risk on their own because knives are your product — they become high-risk only when they co-occur with a firearm term.
Controlled substances
−30 pts / hitWeed, marijuana, cannabis, THC, cocaine, heroin, meth, LSD, MDMA, prescription pills for sale, narcotics. Among the most aggressively filtered categories.
Spam & pressure
−8 pts / hit'Act now', 'buy now', 'limited time', 'last chance', 'today only', 'urgent', 'reply YES', 'you've been selected', 'winner'. Reads like spam to carriers and recipients.
Financial / scam
−18 pts / hit'Guaranteed', 'no risk', 'risk free', 'money back', '100% free', 'wire transfer', 'Western Union', 'bitcoin', 'cash prize'. Common in scam texts; scrutinized by carriers.
Adult content
−26 pts / hitXXX, escort, explicit, nude, hookup, sensual, enhancement. Filtered by most US carriers.
Aggressive language
−16 pts / hit'I'll kill', 'threaten', 'hurt you', 'come after you', 'or else'. Trips both carrier filters and legal review.
What we never flag
knife, cutlery, sharpening, and Cutco are your legitimate business. They are never flagged on their own — only when one of them sits inside a firearm co-occurrence trap.
We flag, we don't censor
The engine surfaces risk and suggests a safer rewrite, but it never blocks you. The final send decision always belongs to the rep. When it suggests a rewrite, it keeps your offer, dates, times, and place exactly as you wrote them — it only neutralizes the risky combination and adds compliant opt-out language.
Opt-out & compliance
Carriers expect promotional texts to carry clear opt-out instructions. A marketing message without "Reply STOP to opt out" is itself a flag — missing it is a compliance risk on top of any content risk. The suggested rewrite always adds it, and adds "Msg & data rates may apply. HELP for help" when there's room.
Two ways to check
A deterministic rules engine that runs entirely in your browser. No data leaves your device, no per-message cost. Best for a quick gut-check before you send.
Sends your draft to an external language model that judges meaning, not just keywords, and writes a tailored safer rewrite. Use it when the wording is subtle or borderline.