Guides
How STIR/SHAKEN, carrier spam labeling, and branded calling affect AI voice agent programs — and a practical compliance checklist to keep your outbound calls reaching the people who asked you to follow up.
Last updated
If your AI voiceAI voiceAn artificially generated, natural-sounding voice produced by a TTS model. Thoughtly supports a library of AI voices and brand-specific cloning. agents are placing outbound calls, the single biggest threat to your contact rateContact rateThe percentage of inbound leads your team actually reaches by phone. Most B2C teams hover around 25%; Thoughtly typically delivers 90%+. is not the script, the offer, or the time of day. It is whether the call reaches the recipient's screen at all — or gets buried under a "Spam Likely" label before the first ring finishes.
CarrierCarrierA telecommunications provider that routes phone calls and SMS over its network. Twilio, Telnyx, and Bandwidth are the three most common in the AI voice space. spam labeling, STIR/SHAKEN call authentication, and branded calling are now table stakes for any team running high-volume outbound voice workflows. Misunderstand how they work, and your answer rates collapse. Get them right, and you build a durable trust layer that keeps your calls reaching the people who asked you to follow up.
This guide covers the regulatory framework, how carrier spam labeling actually works, a practical compliance checklist, and how to use branded calling to protect your outbound voice program.
According to Hiya's 2026 State of the Call report, 86% of unknown calls go unanswered. When a carrier attaches a "Spam Likely" or "Scam Likely" label to your number, you are not just losing a few percentage points of pickup — you are effectively muting your outbound channel.
For teams running AI voice agents on inbound lead follow-up — insurance quotes, mortgage pre-qualifications, appointment confirmations, education enrollment callbacks — a spam label turns a warm lead into a dead one. The prospect filled out your form 30 seconds ago, your agent calls back, and the phone shows "Potential Spam." That is a conversion you paid to acquire and lost before anyone spoke.
The compounding damage goes beyond individual calls:
Several overlapping federal rules govern caller ID, call authentication, and the trust signals carriers use to decide whether a call is legitimate.
STIR (Secure Telephone Identity Revisited) and SHAKEN (Signature-based Handling of Asserted information using toKENs) are a suite of protocols that let carriers verify whether the calling party is authorized to use a given phone number. The FCC mandated STIR/SHAKEN implementation for large voice service providers by June 30, 2021, under the TRACED Act (Telephone Robocall Abuse Criminal Enforcement and Deterrence Act, Pub. L. 116-105, signed December 2019).
STIR/SHAKEN assigns one of three attestation levels to each call:
| Attestation Level | Meaning | Typical Scenario |
|---|---|---|
| A — Full | The carrier verified the caller's identity and their right to use the calling number. | Business calling from a number registered to their account with the carrier. |
| B — Partial | The carrier verified the customer relationship but cannot confirm authorization for the specific number. | Business calling from a ported or third-party number. |
| C — Gateway | The carrier is the network entry point but cannot verify the call's origin. | Calls entering from international gateways or interconnected VoIP providers. |
Attestation level directly affects how downstream carriers and analytics engines treat the call. Full A-level attestation is the strongest trust signal a call can carry. Calls with C-level attestation are far more likely to receive spam labels.
The Truth in Caller ID Act (47 U.S.C. § 227(e), amended by the RAY BAUM'S Act in 2018) prohibits transmitting misleading or inaccurate caller ID information with the intent to defraud, cause harm, or wrongfully obtain anything of value. This applies to voice calls, text messages, and IP-based voice services.
For AI voice teams, this means:
Violations can result in FCC enforcement actions with forfeiture penalties of up to $10,000 per violation, with a maximum of $1,000,000 for continuing violations (47 U.S.C. § 503(b)).
The FCC requires all voice service providers to file certifications in the Robocall Mitigation Database, confirming either full STIR/SHAKEN implementation or a robocall mitigation program. Providers that fail to file — or whose filings are deficient — can have their traffic blocked by downstream carriers. If your voice platform provider or carrier is not in the database with a current filing, your calls may be blocked before they even reach the spam-labeling stage.
Understanding the mechanics of spam labeling helps you avoid triggering it. Carrier spam labels are not just based on STIR/SHAKEN attestation — they layer multiple signals.
Major US carriers partner with third-party analytics companies to score calls:
| Carrier | Analytics Partner / Tool | Label Examples |
|---|---|---|
| T-Mobile | T-Mobile Scam Shield (internal + partners) | "Scam Likely" |
| AT&T | Hiya / AT&T ActiveArmor | "Suspected Spam," "Spam Risk" |
| Verizon | TNS / Verizon Call Filter | "Potential Spam" |
These analytics engines evaluate calls using:
Spam labels are applied automatically by analytics engines when a number's trust score drops below a carrier-specific threshold. Removal is not automatic — you typically need to:
Prevention is significantly easier than remediation.
Use this checklist to protect your AI voice program's caller reputation from day one.
| Step | Action | Why It Matters |
|---|---|---|
| 1 | Verify STIR/SHAKEN A-level attestation with your carrier or platform | Full attestation is the strongest trust signal. Confirm your provider delivers it for your numbers. |
| 2 | Register all outbound numbers with CNAM | A registered business name reduces "Unknown Caller" flags and builds number identity. |
| 3 | Register numbers with Hiya, First Orion, and TNS | Proactive registration with all three major analytics providers prevents mislabeling before it starts. |
| 4 | Enroll in branded calling programs | Displaying your verified business name and/or logo on the recipient's screen improves trust and answer rates. |
| 5 | Warm numbers gradually | Do not send hundreds of calls from a newly provisioned number on day one. Ramp volume over 7–14 days. |
| 6 | Monitor answer rates and short-call ratios | A sudden drop in answer rate or increase in sub-10-second calls is an early indicator of spam labeling. |
| 7 | Maintain consent records and suppression lists | Only call people who have given prior express consent. Scrub against DNC lists and your own suppression list. |
| 8 | Use caller ID numbers you own and can receive callbacks on | Comply with the Truth in Caller ID Act. Never use unregistered or unverifiable numbers. |
| 9 | Rotate numbers responsibly — not to evade detection | Using a number pool is fine for load distribution. Using it to dodge spam labels is a compliance violation and accelerates flagging. |
| 10 | Respond to callback attempts | If a recipient calls back and gets a dead line, that signals illegitimacy to carriers and analytics engines. |
Thoughtly provides several features that support compliant, high-trust outbound voice operations. Each feature referenced below is documented in Thoughtly's current product documentation.
Thoughtly's branded calling feature displays your verified business name on supported carrier networks and devices. When a recipient sees your company name instead of an unknown number, they are significantly more likely to answer.
Branded calling is carrier-dependent — display behavior varies by network and device — but it is one of the most effective tools for improving outbound answer rates on legitimate follow-up calls.
Call screening bypass lets Thoughtly voice agents respond to automated screening prompts (iOS Live Voicemail, Google Call Screen, Truecaller, carrier-level screening) with your business name and reason for calling. Instead of getting silently dropped by a screening filter, the agent identifies itself truthfully and proceeds to the conversation.
Thoughtly's caller ID configuration lets you mask outbound calls to display a specific verified phone number, ensuring calls appear from a recognized central number rather than a rotating pool. The number must be verified through a validation call before it can be used.
Consent mode and suppression lists in Thoughtly help teams maintain compliant outbound workflows by gating calls behind verified consent records and automatically suppressing contacts who have opted out or appear on DNC lists.
Thoughtly's phone number settings include an outbound calling toggle that lets teams immediately stop all outbound calls during a spam incident, number compromise, or compliance investigation — without deleting agents or workflows.
STIR/SHAKEN is a call authentication framework that verifies the caller's right to use a phone number. It operates at the carrier/network level and is invisible to the recipient. Branded calling is a separate trust layer that displays your verified business name and/or logo on the recipient's screen. They solve different problems — STIR/SHAKEN prevents spoofing; branded calling builds visual trust — and they work best together.
Yes, but it requires proactive steps. Register your number with the analytics providers that serve the carrier where the label appeared (Hiya for AT&T, First Orion for T-Mobile, TNS for Verizon). Submit a dispute with documentation showing legitimate use and consent. Improve your calling patterns — reduce volume, ensure consent, avoid short-duration calls. Labels typically age off within days to weeks after patterns improve, but proactive registration accelerates the process.
STIR/SHAKEN applies to all voice calls traversing IP networks, regardless of whether the caller is a human, an IVR, or an AI agent. The protocol authenticates the number, not the entity speaking. If your AI voice agents are placing calls through a carrier that supports STIR/SHAKEN, those calls will receive attestation ratings based on the same criteria as any other call.
There is no published universal threshold. Spam detection depends on a combination of volume, velocity, call duration patterns, complaint rates, attestation level, and number history. The safest approach is to ramp volume gradually (start with 20–50 calls per day per number and increase over 7–14 days), ensure every call is to a consented recipient, and monitor answer rates for signs of labeling.
Branded calling is not required by federal law. It is a voluntary trust layer offered through carrier partnerships and analytics providers. However, the competitive reality is that teams without branded calling experience materially lower answer rates, which translates directly to lost revenue on leads they already paid to acquire.
This article is informational and does not constitute legal advice. Consult qualified legal counsel for compliance decisions specific to your organization.
