STIR/SHAKEN and spam-likely labeling

Telephony & InfrastructureSTIR/SHAKENalso: STIR/SHAKEN and spam-likely labellingalso: spam likely labelalso: scam likely labelalso: caller ID authentication

In one sentence

STIR/SHAKEN is a framework that lets carriers cryptographically verify a call really originates from the number it claims, while spam-likely labeling is a separate and proprietary scoring layer that decides whether a warning appears on the recipient screen.

A, B, C STIR/SHAKEN framework specification, 2026Last reviewed 30 July 2026

Not to be confused with TCPA, or Voice biometrics and anti-spoofing.

Definition

STIR/SHAKEN lets phone carriers check, using a cryptographic signature, that a call is genuinely coming from the number on the screen.

Spam-likely labeling is a different thing entirely. It is the warning a phone company puts on the screen when it guesses the call is a nuisance.

STIR stands for Secure Telephony Identity Revisited and SHAKEN for Signature-based Handling of Asserted information using toKENs. Together they provide caller identity authentication across IP-based voice networks. The carrier the call originates on signs it with a certificate asserting its relationship to the calling number, and carriers downstream verify that signature.

The three attestation levels, which is the part that matters operationally

  • A, full attestation. The carrier knows the customer and has verified their right to use that number.
  • B, partial attestation. The carrier knows the customer but has not verified the number.
  • C, gateway attestation. The carrier is only passing the call along and can vouch for nothing.
  • The level materially influences whether a call is delivered cleanly, labeled, or blocked. Reaching A requires a proper carrier relationship with verified number ownership, which is a commercial arrangement rather than a configuration setting.

What it does and does not establish

  • It authenticates that the calling number is being used legitimately.
  • It does not verify that the call is wanted, honest, or anything other than a scam.
  • A fully attested call from a properly registered number can still be a fraud attempt. Authentication is not reputation.

Spam labeling is a separate and far less transparent system

  • Analytics providers, working with carriers and handset makers, score numbers and apply labels such as spam likely or scam likely.
  • The inputs include call volume, answer rate, average duration, complaint reports and patterns of very short calls.
  • The scoring is proprietary, differs between providers and between carriers, and in most cases is not subject to any meaningful appeal.
  • A legitimate business can be labeled on pattern alone. A high volume of short unanswered outbound calls looks like spam whether or not it is.

What this means for an outbound voice agent

  • Labeling collapses answer rates, and a collapsed answer rate destroys the economics of an outbound campaign.
  • Remediation means registering numbers with analytics providers, joining branded calling programs, pacing calls sensibly, securing the right attestation level, and staying away from the patterns that trigger scoring in the first place.
  • Rotating through many numbers to dodge a label is a recognized abuse pattern. It is increasingly detected, and it puts a business on the wrong side of the system it needs to work with.

Common misconception

That STIR/SHAKEN prevents spam. It authenticates identity, nothing more. Labeling is a separate, opaque, analytics-driven layer sitting on top of it. Running the two together leads businesses to assume that compliance guarantees deliverability, and it does not.

Why it matters commercially

For an outbound voice agent this is an operational threat rather than a compliance footnote. A labeled number means nobody picks up, and an agent nobody picks up to has nothing to do. It is also the strongest single argument for inbound and on-site voice over outbound calling, because neither of those meets the problem at all.

In voice specifically

Automation makes the trigger patterns worse rather than better. An outbound agent can dial faster than a team of people, which raises volume and shortens average duration at the same time, and those are two of the inputs the scoring watches. The technology that makes outbound cheap is the technology that makes it look like spam.

Where AsqVox fits

None of this reaches a browser conversation. Voice that runs on the website through a single script tag involves no outbound calling, no phone number, no carrier and no attestation, so there is nothing to sign and nothing to label. It is a structural property of where the conversation happens rather than a feature, and it is worth stating plainly in comparisons, because buyers researching voice agents have usually already met a deliverability problem.

Visual

Two systems, often confused

Two systems, often confusedSTIR/SHAKEN: identityOriginating carrier signs, transit, terminating carrier verifiesAttestation A, full: thecarrier knows the customer andhas verified their right tothe numberAttestation B, partial: thecarrier knows the customer buthas not verified the numberAttestation C, gateway: thecarrier is only passing thecall along and can vouch fornothingProves the number is beingused legitimatelyDoes NOT prove the call iswantedFull attestation does notprevent labeling.Spam labeling: reputationA separate analytics layer scoring the number, ending on the handset as SPAM LIKELYInputs: call volume, answerrate, average duration,complaint reportsOutput: a SPAM LIKELY warningon the recipient screenProprietary scoring, varies bycarrier, limited appealWhat triggers a label: highvolume outbound, low answerrate, very short calls, numberrotation

Authentication is not reputation. You can be verified and still be ignored.

Not applicable to on-site browser voice: no number, no carrier, no label. The two halves are run by different parties, measure different things, and neither one can fix the other.

Statistics

Every figure carries its source and year. Vendor numbers are labelled as vendor numbers, and where no reliable figure exists this page says so rather than borrowing one.

STIR/SHAKEN attestation operates at three levels: A full, B partial and C gateway, reflecting how much the originating carrier knows about the customer and whether it has verified ownership of the number.

A, B, Cindustry range

STIR/SHAKEN framework specification, 2026 - A framework fact rather than a measurement, and it is the one operational detail worth memorizing. Attestation level materially influences whether a call is delivered cleanly, labeled or blocked.

In the United States the FCC required major voice service providers to implement STIR/SHAKEN in their IP networks, with the principal compliance deadline on 30 June 2021 and extensions applying to certain smaller providers.

30 June 2021industry range

US Federal Communications Commission, 2021 - Implementation status and equivalent frameworks vary by country. Check the current national regulatory position for the market you are calling into rather than assuming global coverage.

Spam labeling algorithms are proprietary. No threshold is published, no scoring model is disclosed, and no reliable statistic exists for how often legitimate businesses are labeled in error.

-no reliable figure

The opacity is itself the most important fact on this page. A business cannot audit the system that decides whether its calls are answered, which is why remediation is slow and why prevention beats appeal.

No authoritative public benchmark exists for the answer-rate impact of spam labeling. Figures circulate in vendor material and are not independently verified.

-no reliable figure

Anyone quoting a precise percentage drop is quoting their own telemetry. The direction is not in dispute. The size of it is not established.

Deepfake fraud attempts rose by more than 1,300 percent during 2024, from an average of one per month to seven per day, across a sample of more than 1.2 billion analyzed calls.

Over 1,300%vendor claim

Pindrop, 2025 Voice Intelligence and Security Report, 2025 - Vendor telemetry drawn from Pindrop's own sample, so cite it as such. It is adjacent context for why caller identity is under pressure, not evidence about labeling.

Generative AI was projected to push US fraud losses to USD 40 billion by 2027, up from USD 12.3 billion in 2023, a compound annual growth rate of 32 percent.

USD 40bn by 2027analyst forecast

Deloitte Center for Financial Services, 2024 - An analyst projection, and it should be cited with its date. It explains the pressure the identity layer is under. It says nothing about whether any particular number gets labeled.

Examples

In practice

An outbound appointment-reminder agent watches answer rates fall sharply over three weeks. Investigation finds two causes at once. The calls carry C-level gateway attestation, because they originate through a reseller with no verified number ownership. And the pattern of high-volume short calls has tripped spam scoring at two major carriers. Remediation runs for weeks: establish a direct carrier relationship to get A attestation, register the numbers with the analytics providers, cut the pacing, and lengthen the opening so answered calls stop terminating after three seconds.

The everyday version

This is why people stopped answering unknown numbers. Phone networks now try to check that callers are who they say they are, and separately, phone companies guess which calls are nuisance calls and put a warning on the screen. If your business picks up that warning, your calls stop being answered, and getting it removed is slow and frustrating. None of it applies if the conversation happens on your own website.

Usage

Who says it

  • Telephony and compliance specialists, and outbound campaign operators, who live with it daily.
  • Contact center operations teams, who meet it as a deliverability problem rather than a security one.
  • Rarely anyone working on inbound or on-site voice, where it does not apply.

Where it turns up

  • Usually in the same clause as outbound calling capability, attestation level, number registration and branded calling, call pacing controls, and consent and TCPA compliance.
  • In US outbound procurement it is a standard question. A vendor that cannot speak to attestation level is usually excluded on the spot.

Common misuse

  • Assuming STIR/SHAKEN compliance guarantees delivery without a label. It does not, and this is the most consequential misunderstanding in outbound voice.
  • Rotating numbers to escape labels. It is detected, it is penalized, and it reads badly to anyone who later audits the account.
  • Applying outbound deliverability worries to inbound or on-site voice, where they are structurally absent.

Questions people ask

Does STIR/SHAKEN stop spam calls?

No. It authenticates that the calling number is being used legitimately by the party placing the call. It does not verify that the call is wanted, honest or anything other than a scam. A fully attested call from a properly registered number can still be a fraud attempt, because authentication is not reputation.

What are attestation levels A, B and C?

A is full attestation: the carrier knows the customer and has verified their right to use the number. B is partial: the carrier knows the customer but has not verified the number. C is gateway: the carrier is only passing the call along and can vouch for nothing. The level materially influences whether a call is delivered cleanly, labeled or blocked, and reaching A requires a proper carrier relationship with verified number ownership.

Why is my business number showing as spam likely?

Because analytics providers working with carriers and handset makers score numbers on call volume, answer rate, average duration, complaint reports and short-call patterns. A high volume of short unanswered outbound calls looks like spam whether or not it is. The scoring is proprietary, varies between carriers, and in most cases carries no meaningful appeal, so remediation is slow.

Does STIR/SHAKEN apply to a voice agent on a website?

No. A conversation held in the browser has no phone number, no carrier and no attestation, so there is nothing to sign and nothing to label. That is a structural difference between on-site voice and outbound calling, and it is the strongest single operational argument for inbound and on-site over outbound.

Share this definition

Last reviewed 30 July 2026. Written and reviewed by Dhruv Dholakia, founder of AsqVox.