The 10DLC Regime

1What 10DLC is, structurally

10DLC ("10-digit long code") is the compliance regime governing application-to-person SMS/MMS sent from ordinary ten-digit phone numbers to US mobile subscribers. In practice it is a licensing system for the right to send text messages at scale, operated by a private consortium rather than a public regulator.

The system has three properties that shape everything else about it:

• Registration is mandatory in effect. Unregistered A2P traffic on long codes is filtered or blocked by the carriers. There is no unregistered path to deliverability at volume.
• The rules are unpublished. There is no specification document, no changelog, and no public rule index. The operative rules live in portal validation code, in PDFs behind authenticated logins, in per-account API responses, and in the private judgment of two vetting vendors and the carriers' review teams.
• Errors are billed. Failed submissions, failed vets, and appeals are charged as fees regardless of outcome (§12). The absence of a published specification and the per-error fee model are the same fact viewed from two sides.

2Evidence tiers & method

Every rule in this document carries an implicit evidence tier. Conflating them is the most common analytical error in this domain, because frontend validation is evidence of the operator's intent, not proof of server behavior — API submissions bypass the portal UI entirely, and the server is in places looser (fields the wizard demands that the API accepts empty) and in places stricter (duplicate caps and name-obfuscation checks that exist only server-side, surfaced as numbered error codes).

Method used to compile this reference: recovery of the portal's original source tree from published sourcemaps (tier 3); capture of live enumeration endpoints from an authenticated registry account (tier 2); reading of the operator's own policy corpus (tier 2); and analysis of a large body of real registration and campaign outcomes, including verbatim rejection text with numeric codes (tier 1). Where tiers disagree, tier 1 governs — and §9 documents several places where the live evidence corrects what the frontend implies.

3Actors and the chain of custody

A message's right to exist passes through five distinct parties, each of which can reject it:

Live enumeration returns seven US networks, each with a registry network ID and an osrBitmaskIndex used for per-carrier status packing:

4The brand rulebook

A brand is the registered legal identity of the sender. It is the durable object; campaigns attach to it. Brand registration costs a non-refundable fee (§12) and any subsequent change to key identity fields costs the same fee again.

4.1 Field rules

Tier 3 unless marked. Lengths and regexes are as the portal enforces them.

4.2 Entity types and their conditional matrix

Live enumeration returns exactly five: PRIVATE_PROFIT, PUBLIC_PROFIT, NON_PROFIT, GOVERNMENT, SOLE_PROPRIETOR.

Cross-field mechanics the portal enforces on edit: changing entity type clears the public-company fields; changing country clears the entire address and the tax ID; changing the tax-ID issuing country clears the tax ID. Sanctioned jurisdictions — Cuba, Iran, North Korea, Syria — are excluded from every country selector.

4.3 Verticals

Twenty-three, each with a stable key, an industry slug, a display name, and a description. Live keys: AGRICULTURE, COMMUNICATION, CONSTRUCTION, EDUCATION, ENERGY, ENTERTAINMENT, FINANCIAL, GAMBLING, GOVERNMENT, HEALTHCARE, HOSPITALITY, HUMAN_RESOURCES, INSURANCE, LEGAL, MANUFACTURING, NGO, POLITICAL, POSTAL, PROFESSIONAL, REAL_ESTATE, RETAIL, TECHNOLOGY, TRANSPORTATION.

5Identity verification & vetting

5.1 The identity status ladder

Four values, in ascending order of trust: SELF_DECLARED → UNVERIFIED → VERIFIED → VETTED_VERIFIED. Registration submits identity facts; the registry's verification partner checks them against external registries; the result lands asynchronously.

5.2 Vetting classes

Live enumeration, with validity periods — note that all five report a 1200-month validity (effectively perpetual) at the registry layer, which is not the same as the vendor-side or election-cycle expiries described in policy documents (§6.2):

5.3 What the score actually measures

The vetting score is a 0–100 identity-confidence number. Analysis of a large outcome corpus establishes three things about it:

• Roughly 90% of its variance is explained by factors outside the submission — business registries, web-domain authenticity and age, firmographic data. The payload is a minority contributor. This is a bureau product, not an assessment of the message program.
• The in-payload levers that do move it are coherence signals. Contact email on the company's own domain rather than a free mailbox: on the order of +3 points average, with elite (70+) outcomes several times more likely. A merchant-owned website rather than a platform-issued link: on the order of +4 points. Short DBAs closely derived from the legal name: the largest single in-payload weight.
• Graders disagree materially. The two commercial vetting vendors score the same population about eleven points apart at the median. A score is therefore never a fact on its own; it is a (score, vendor, epoch) tuple, and any system that stores or acts on scores without the vendor and date is calibrated against noise.

The practical implication for anyone building on this: do not gate on a predicted score. Use score drivers as coaching, because the ceiling of payload optimization is low and the vendor variance is high. Do gate on the mechanical rules, where the ceiling is 100%.

5.4 Authentication+ (public companies)

A distinct verification product with unusually sharp edges. It verifies two things: that the brand's registered web domain matches the business-contact email's domain, and that a human at that address completes a 2FA attestation.

• Prerequisites: public-company entity type; business contact email populated; identity already VERIFIED or better; no other pending Auth+ vet. Distinct numeric errors are returned for a missing business email, a wrong identity status, and a non-public entity type.
• The attestation PIN is valid 7 days; resend is permitted while pending. Failure to complete within 30 days fails the vet non-appealably — the remedy is buying another vet.
• Once active, brand identity fields cannot be changed at all; a new brand is required. Changing the business contact email is permitted but expires the active vet, blocking new campaign registration until re-verification.
• Appeals are limited to failed results, within 45 days, under two categories: email-ownership and domain-ownership.

The generalizable lesson from Auth+ applies to every entity type, because it reveals what the verification layer actually checks: the contact email's domain is compared to the registered website's domain. Coherence between those two fields is a real, mechanized check — not a heuristic.

6The 27 use cases

A use case declares what a campaign is for. It determines the monthly fee, the carrier throughput class, the eligibility prerequisites, and — least documented and most consequential — the scope of content the campaign may carry.

6.1 Where use-case rules actually live

There is no published table mapping use cases to their requirements, because the mapping is distributed across three mechanisms:

6.2 Eligibility gates by tier

Three distinct gate types recur across the catalogue:

• Post-registration carrier approval required — the campaign registers, then waits for a carrier decision: carrier-exempt, conversational, emergency, sweepstake, agents-and-franchises, proxy, K-12 education.
• Restricted to pre-approved submitters — the CSP itself must be approved before the use case appears: social, platform free trial; and by separate carrier-reviewed application, both UCaaS tiers.
• Entity or credential gated — charity (verified 501(c)(3)); political (non-profit legal form plus either a verified 501(c)(3/4/5/6) subsection or a political vet); sole proprietor (an EIN-less individual); public-safety (not available to the public at all).

Political deserves its own note because its enforcement mechanism is temporal rather than textual: political vets expire on dates that vary by election type, and on expiry all of the brand's political campaigns deactivate within 24–48 hours. If the expired vet was also what confirmed the brand's tax-exempt status, the brand drops to UNVERIFIED and can register nothing at all. Political tokens are single-use; the non-billable federal provider can issue several per vet, the billable one cannot.

6.3 The complete catalogue, with sub-use-case structure

Tier 2 — live enumeration. "Sub" columns are the minimum and maximum child use cases; "valid child" marks a use case that may itself be selected as a child of another.

Note the asymmetry: of the thirteen standard use cases, ten may serve as children of a MIXED or LOW_VOLUME parent; MIXED, LOW_VOLUME, and M2M may not. No special use case may be a child.

6.4 Use-case denials by entity type

Verified government brands may not declare charity, political, sole-proprietor, or platform-free-trial use cases on either major carrier, and additionally not carrier-exempt on one of them. The structural reason is that each of those has its own entity gate incompatible with government status — which illustrates a general principle: use-case eligibility and entity type are jointly constrained, and the intersection is computed server-side rather than published.

7The campaign rulebook

A campaign is a registered message program attached to a verified brand. It carries the copy, the consent mechanics, the declared attributes, and the carrier provisioning state.

7.1 Copy bounds

7.2 The eight declared attributes

Every campaign must answer eight booleans explicitly — no defaults are accepted: subscriber opt-in, subscriber opt-out, subscriber help, embedded link, embedded phone, age-gated, direct lending, and terms acceptance. Three mechanics govern them:

• Carrier metadata can force values. Where a selected carrier requires subscriber opt-in/opt-out/help, the toggle is locked to yes.
• Terms acceptance is effectively mandatory — locked to yes in the current form, asserting the campaign follows industry messaging principles. Submitting it false is a documented registry failure (§9.3).
• Age-gating, direct lending, and number pooling are declarations only — no additional fields are collected. Number pooling additionally signals an intent to use fifty or more numbers, which triggers a different provisioning path on one carrier.

7.3 Keyword mechanics

Opt-out and help keywords default to STOP and HELP. Characters are restricted to letters, digits, and spaces; entries are auto-uppercased; the aggregate joined length is capped at 255 with no per-keyword count limit. Critically, the defaults are deletable — no client rule forces STOP or HELP to remain, so that obligation is enforced only downstream by carrier review (and §9 shows it is).

8The copy specification

This is the least documented and most frequently violated part of the regime. The requirements below originate as helper prose in the submission portal — nominally tier 3 — but §9.1 reproduces the carriers' own verbatim rejection text enforcing them almost word for word, which makes this section tier 1 in effect.

8.1 The three consent messages

8.2 The six call-to-action elements

The CTA must describe how consumers opt in and grant consent, be explicitly clear about the nature of the program, and list every opt-in method if more than one exists. It must display: (1) the program name, (2) message frequency, (3) the "message and data rates may apply" disclosure, (4) a STOP keyword, (5) a link to terms and conditions, (6) a link to a privacy policy.

8.3 Hosted document requirements

• Terms and conditions must contain: message frequency, a description of the message program, fees, opt-in and opt-out instructions, customer-care contact information, and a reference to the privacy policy. It must be consistent with the registered brand.
• Privacy policy must explain how personal data is collected and used, and must state that mobile opt-in data is not shared with third parties for marketing purposes. It too must be brand-consistent.
• Both must actually resolve. "Brand-consistent" in practice means hosted on a domain that reads as the brand's own — see the website-verification failure codes in §9.2.

Upload channels are content-typed and mutually exclusive: the supporting-document slot is for CTA, terms, and privacy evidence and explicitly not for message samples; the sample-multimedia slot is explicitly not for CTA or policy material. Five files each, ten megabytes apiece.

9Rejection taxonomy & carrier codes

Since a 2025 platform release, a party declining a shared campaign must select at least one category and provide an explanation — which made rejection reasons machine-readable for the first time. Nine categories exist at the registry layer; beneath them the carriers use their own numeric codes, which are not published anywhere and are reproduced here from observed rejections.

9.1 The nine registry categories

9.2 Observed carrier codes Tier 1

These numeric codes appear in real rejection payloads. They are the operative vocabulary of the review layer and are documented in no public source.

9.3 Verbatim server and carrier text Tier 1

These strings are the specification, as the enforcing systems state it.

Consent-message content (code 611)

EIN format

Business email quality

Registration coherence (code 603)

Website authenticity (codes 804 / 807)

Other observed registry failures

Note that the 611 text is materially stricter than the portal's own helper prose: the portal says the brand name "may" appear in the opt-out and help replies, while the carrier requires it. Where tier 3 and tier 1 diverge, this is the direction the divergence takes — the enforcing layer is harsher than the submitting layer suggests.

10Carrier terms & throughput

Approval is not binary; it comes with a throughput grant that varies by carrier, use case, entity type, and vetting outcome. The two largest carriers use different models:

• One carrier assigns a message class per campaign — a letter grade mapping to messages per minute, separately for SMS and MMS. Government entities and certain special use cases receive fixed classes that vetting cannot improve.
• The other assigns a brand tier and a daily cap, where the cap is a brand-level property rather than per-campaign. Verified non-profits and government entities receive elevated or uncapped treatment; a political vet can lift an otherwise-capped brand.

Two structural consequences follow. First, throughput is a function of identity, not of need — the way to send more is to be more verifiable, not to ask. Second, some grants are vet-immune: for government entities and several special use cases, buying a vet cannot improve terms, so vetting spend there is pure waste. Sole-proprietor and free-trial programs carry hard per-number daily caps that the submitting CSP is contractually responsible for enforcing.

11Lifecycle, events & traps

11.1 The state machines

Brand: created → identity verification (async) → verified or failed → optionally vetted → (on identity edit) demoted to unverified. Campaign: created → shared upstream → accepted or declined by each connectivity partner → provisioned or rejected per carrier → live → renewed monthly → optionally suspended → expired or deactivated.

Observed live campaign states include registry-side failure, carrier rejection, full carrier provisioning, expiry, and per-carrier suspension flags — the last of which can be set on an otherwise healthy campaign.

11.2 The event stream

The registry emits an immutable event stream covering brand lifecycle (creation, identity-status change, identity-vet failure, appeals, revets, one-time-password verification), campaign lifecycle (creation, billing, updates, expiry), the sharing chain (shared, accepted, declined — carrying the mandatory rejection categories), per-carrier operations (review, approved, rejected, suspended, unsuspended), aggregator-level suspensions with a category, complaints, and carrier suspension rules. Billing events fire on creation and on every renewal, which makes the stream a usable cost telemetry source.

11.3 Four traps worth institutional memory

1. The exclusion deadlock. A downstream aggregator may exclude specific carriers when accepting a campaign. If an excluded carrier is not explicitly removed by the submitter, the "fully provisioned" completion event never fires. Any process waiting on that event waits forever; the only exit is an explicit removal call.
2. EIN-keyed suspension. Carrier suspension rules may key on the brand identifier or on the tax ID. An EIN-keyed rule suspends every brand sharing that tax ID on that carrier at once. Any entity that registers many brands under one tax ID carries correlated, fleet-wide risk.
3. Silent identity demotion. Covered in §5.1 — the drop to unverified is announced only by an event, and blocks new registration while leaving existing campaigns running. A monitor that only watches campaigns will not see it.
4. Asymmetric nudge routing. A formal unstick mechanism exists, with counterintuitive routing: a review nudge may be sent only while a campaign is pending and goes to the last party that has not accepted; an appeal nudge may be sent only after rejection and goes to the last party that did accept — because the rejector has left the chain.

12The fee schedule

Prices as published in the operator's own fee document. The pattern to notice is not the magnitude but the structure: failure is billed, appeals are billed regardless of outcome, and every correction to a mistake costs the price of the original action.

The economics in one sentence: the regime charges per error against rules it does not publish. Any party that reconstructs the mechanical layer and validates before submitting converts a recurring, unpredictable tax into a fixed engineering cost — which is the entire practical argument for documents like this one.

13Special programs

13.1 Sole proprietor

A path for entities without an EIN — explicitly not a cheap alternative for EIN-holders, and audited as such. Limits: one campaign and one phone number per brand. SMS one-time-password verification is mandatory before any campaign may register, and re-verification is required after a mobile-number change. Cross-submitter duplicate caps apply and are enforced server-side with distinct numeric errors: three brands per mobile number, ten per email address, ten per street address — across all submitters industry-wide. Disposable email domains are refused; names are run through an obfuscation detector. Monthly per-carrier volume reporting is mandatory even for zero traffic, with a five-business-day cure period before the submitter's entire sole-proprietor book is disabled.

13.2 Platform free trial

For non-paying, time-limited trial users. Restricted to pre-approved submitters under contract; the trial brand is created by the registry itself, not by the submitter. One campaign for the submitter plus one per reseller. One verified sending number per participant; a maximum of ten opted-in destination numbers per participant; a hard per-number daily cap the submitter must enforce; and a mandatory watermark on every message naming the submitting platform. Monthly compliance reporting required. Vetting cannot improve free-trial terms.

13.3 Government

US-based with a valid EIN, verified against the tax authority; foreign government bodies are ineligible and must register as private companies. Enhanced vetting is never required. Verified status grants elevated fixed carrier terms that vetting cannot further improve, and bars several use cases (§6.4).

13.4 Non-profit edge cases

Where automated verification returns no tax-exempt subsection — common for small religious organizations absent from tax-authority databases — a paper attestation path exists: an identity-status or vet appeal accompanied by a sworn form asserting exclusively charitable purpose, no private inurement, no substantial lobbying, no political-campaign intervention, and lawful purposes. All five assertions are required. Note the content implication: an organization cleared by this route has formally attested away political-intervention messaging.

13.5 UCaaS and RCS

UCaaS requires a carrier-reviewed application by the submitter, and mandates that each business using the platform registers as its own brand — aggregating businesses under one submitter brand is prohibited. Description and samples must name the specific business and its intended use. Automated or API-driven traffic is excluded, as is registering UCaaS traffic under any other use case.

RCS (rich messaging) is an adjacent regime with its own vet class, valid one year, and its own gates: sole proprietors excluded entirely; brand logo and banner each individually verified for a fee; a platform-minted agent identifier per campaign; external content verification after carrier acceptance, initially without appeal; and asset immutability after verification. A successful RCS vet also confers Auth+ compliance.

14Posture versus practice

The regime's published content rules and its observed enforcement differ, and any accurate reading of the rulebook must account for the gap.

What the documents say: a disallowed-content category names gambling, drugs, hate speech, illegal activities, and high-risk services as strictly prohibited. Carrier codes exist for controlled substances and for age-gating failures. A suspension-rule category references controlled substances.

What the outcomes show: regulated-vertical brands register successfully under ordinary verticals, achieve high identity-vetting scores, and operate provisioned campaigns at scale. The taxonomy contains no vertical for such industries — they are not modelled as a category at all. Enforcement, when it occurs, lands on campaign content and coherence rather than on the brand's industry: the observed prohibited-content rejections are a small minority of submissions, and sit alongside far more numerous consent-copy and website-verification failures.

15Structural conclusions

1. Most of the regime is arithmetic. The mechanical layer — formats, bounds, enumerations, required-field matrices, consent-copy elements — is deterministic and now written down. Analysis of a large outcome corpus puts roughly sixty percent of historical failure volume in this layer, with the single largest cause being a message-content rule that a string check catches. The expertise the ecosystem prices as judgment is mostly an unpublished specification.
2. The judgment layer measures coherence, not merit. Does the DBA derive from the legal name; does the contact email's domain match the registered website; do the campaign's policy pages live on the brand's own domain; does the copy name the registered brand. Four separate carrier codes exist for website verification alone. A brand that is internally consistent and externally verifiable passes; one that is fragmented does not, regardless of legitimacy.
3. Scores are not facts. Ninety percent of vetting variance lies outside the payload, and the two graders differ by roughly eleven points at the median on the same population. Any use of a score must carry its vendor and epoch, and no process should gate on a predicted score.
4. Enforcement is post-hoc as well as pre-hoc. Declared attributes are checked against live traffic; suspensions can key on a tax ID and take out an entire fleet; identity demotion is silent and blocks new registration while leaving existing traffic running. A compliance posture that ends at submission is incomplete by design.
5. Secrecy is load-bearing. Failure is billed, appeals are billed regardless of outcome, and every correction costs the price of the original action — against rules that are not published. The absence of a specification and the per-error fee model are structurally the same fact.
6. The knowledge decays. The regime drifts without a changelog; the operator's own corpus records a mid-2026 tightening. A reconstruction is a photograph. The one durable source of truth is the regime's own output: mandatory rejection categories, carrier codes, vetting scores, and status transitions, captured continuously. A party that structurally records those outputs holds a self-correcting model of an unpublished regulator; a party that does not holds folklore.