Email taught a generation of marketers a comfortable lie: that a message you send is a message that arrives, minus some spam-folder friction you can fix with better subject lines. SMS does not work that way, and in 2026 it works less that way than ever. When you hand a text message to a US carrier, you are not requesting delivery. You are submitting an application. The carrier compares your live message against a registration record you filed months ago, checks your sending identity against a registry, evaluates your traffic pattern against a fraud model, and then decides whether the message exists.

Most businesses discover this the way you'd expect: silently. There is no bounce. There is no error. The API returns a 200, the message shows as sent, and the customer never gets it. I have watched teams spend a quarter debugging their application code for a problem that lived entirely inside a carrier's filtering layer.

The rules got teeth in 2025 and the teeth got sharper in 2026

The turning point was February 2025, when the major US carriers stopped throttling unregistered 10DLC traffic and started blocking it outright — 100% of it, with penalties reported up to $10,000 per violation, and "snowshoeing" (rotating numbers to dodge detection) explicitly named as prohibited rather than merely frowned upon. That flipped registration from a best practice into a precondition for the channel functioning at all.

Then the escape hatch closed. For years the standard workaround was to move traffic to toll-free numbers, which didn't carry the 10DLC registration burden. Carriers anticipated that migration and introduced toll-free verification requirements starting in late 2023. As of January 1, 2026, toll-free A2P messaging is reviewed at essentially the same level as 10DLC: new registrations must supply a Business Registration Number (a US EIN or equivalent government-issued ID), the issuing country, and verified business identity matching official records, including entity type. Already-verified toll-free numbers are grandfathered — the new bar applies to registrations and verification requests submitted after that date. If you were planning to spin up fresh toll-free numbers for a campaign next quarter, that plan just got a paperwork dependency.

Layered on top: T-Mobile enforces daily message caps at the brand level, AT&T enforces per-minute throughput at the campaign level, and sole proprietors are capped hard (reported at 1,000 messages/day on T-Mobile, 15 messages/minute on AT&T). The Campaign Registry raised brand registration to $4.50 and standard brand vetting to $41.50 in August 2025. And the FCC's one-to-one consent rule — vacated once, rescheduled repeatedly — took effect January 27, 2026, meaning consent captured through a shared lead form no longer covers every buyer on that form. States are diverging too: Texas SB 140 pulled texts into "telephone solicitation" with DTPA exposure, and Virginia SB 1339 requires honoring text opt-outs for ten years.

The part engineers keep missing: registration is not a one-time gate

Here is my actual opinion, and it's the reason I wanted to write this. Most teams model carrier registration as a setup task — a form you fill out during onboarding, a ticket you close, a thing that is "done." It isn't. It's a continuously evaluated runtime contract, and treating it as a static gate is the root cause of nearly every mystery deliverability incident I've seen.

Carriers now run real-time matching of your live message content against the sample messages you registered. Drift is a filtering trigger. That matters enormously right now, because the same industry that is racing to put generative models in front of customers is generating SMS copy dynamically — and a model that paraphrases your approved sample into something friendlier has, from the carrier's perspective, sent an unregistered message. High opt-out rates trigger filtering. Unusual volume patterns trigger filtering. Campaigns registered in 2023 and 2024 against older requirements can trigger rejection when re-verified. Vague use-case descriptions get bounced at registration — "customer notifications and updates" is not a use case, it's a shrug — and carriers now want live, reachable opt-in URLs rather than screenshots.

If you internalize one thing: your deliverability is a function of the agreement between what you registered and what you actually send, measured continuously. That is an engineering invariant, and invariants need monitoring, not a checklist.

Why this belongs in the same platform as your calls

The structural mistake most companies make is buying messaging from a vendor that has no idea a phone call exists. You end up with a texting silo: its own numbers, its own opt-out list, its own reporting, and no way to answer whether the customer who texted at 4pm is the customer who called at 10am from a paid search ad.

At Dial800 we built TextingHub the other way around — text-enable the numbers you already own and already advertise, whether toll-free, local, VoIP, or Microsoft Teams. Same number on the truck, same number in the ad, one identity. Toll-free verification and 10DLC brand and campaign registration are both submitted from inside the platform rather than through a separate registry portal, because the handoff between "our tool" and "their portal" is exactly where messaging programs stall out and quietly fail for six weeks. STOP handling runs against a managed opt-out list automatically, and the compliance library covers SHAFT content categories, sender-ID guidelines, and frequency guidance — the parts of carrier filtering that no amount of clean registration will save you from if your content or cadence is wrong.

The part I care most about is the measurement. Text conversations, message counts, and text revenue metrics sit alongside call volume in the same dashboards, and real-time SMS webhooks fire on new conversations and messages so your CRM or warehouse reacts immediately. Inside the contact center, text threads land in the same agent console as calls, with claim/transfer/release, supervisor takeover, a 72-hour reply window, and SMS activity reporting on conversations handled and average first reply time. Voice and messaging stop being two channels you correlate after the fact and start being one customer record.

The test

Ask your messaging provider one question: what happens when a message I send drifts from the sample I registered? If the answer involves you finding out from a customer complaint, you don't have a messaging platform. You have an API that returns 200 and hopes for the best.