Carrier filtering is the blocking or silent dropping of SMS by an operator's anti-spam systems based on content patterns, sender behaviour or volume, often reported as a specific failure reason or not reported at all.
Operators run filters, often called SMS firewalls, on the application traffic entering their networks. They look at content (link shorteners, phishing patterns, banned words), at the sender (unregistered names, numbers with a poor history) and at behaviour (volume spikes, repeated identical texts, traffic arriving on an unexpected route). A filtered message may fail with a reason or, on some networks, be reported delivered and never shown. The rules are local, mostly unpublished, and change without notice.
Registration regimes add their own layer. In the United States, carriers filter application traffic from ordinary 10-digit numbers that is not registered for 10DLC; in India, commercial messages have to match a header and template registered on DLT. Elsewhere the rules are less formal but just as real, and every smishing wave makes filtering stricter for everyone.
Most filtering is avoidable. Hypothetically, a retailer's promotion with a public link shortener starts failing as filtered on one operator; moving the link to the retailer's own domain is the first change to test, on a handset on that network. The habits that help: links on your own domain, codes without links, the brand in the first words, the sender type the destination expects, no marketing sent as transactional, and volume built up gradually on new destinations. The guide to marketing SMS covers content in more detail.
Three misconceptions are worth dropping:
- That delivered proves nothing was filtered. Some networks drop silently.
- That changing one word is enough. Filters weigh the sender and the pattern as well as the words, and each resend is billed.
- That one test phone tells the whole story. Filtering differs by operator.
At SMSMeteor: filtered failures are reported among the plain reasons, with the raw carrier code, where the operator gives one. Before sending, text that matches the acceptable use blocklist (phishing patterns, impersonation of banks, carriers and public services) is rejected with content_blocked, and under the acceptable use policy promotional content flagged as transactional is treated as abuse. When a carrier refuses a sender name, the message falls back to a numeric sender and the report says so.
