Your business number shows as Spam Likely because one of three private analytics companies scored its calling pattern as spam, not because a regulator ruled on it. First Orion labels calls for T-Mobile, Hiya for AT&T and TNS for Verizon, each one scores every call on its own, and the FCC decided in December 2020 not to give callers any right of appeal over a label. Registering your numbers tells those three companies who you are. Only your calling pattern changes the score.
I build outbound AI callers for real estate teams, agencies and lenders, and the label is the first thing a client blames when answer rates fall. I have built outbound systems for 200+ businesses, and the fix is almost never the one the calling platform sells. So I read the rule, the FCC order behind it, the registry's own terms and the analytics companies' help pages.
No regulator put Spam Likely on your number. No regulator will take it off.
The mistake: rotating numbers to outrun the label
The standard advice in AI calling is to spread a campaign across a pool of numbers and swap them as they burn. Retell's guide to avoiding Spam Likely, updated March 1, 2026, tells you to rotate numbers for large campaigns. Bland's custom dialing feature cycles through your number inventory and says it guarantees the highest pickup rate. HighLevel sells local presence dialing, which picks a caller ID that matches the prospect's area code. Every one of these trains the exact signal the label is built on.
The companies that assign the label say so in writing. Hiya's reputation report card grades a number on Maturity, and its own definition of that grade is whether you use established numbers without rotating. First Orion's help center tells callers to avoid rotating numbers, because frequently changing or reactivating dormant numbers looks suspicious to carrier analytics. Telnyx, in a June 5, 2026 article on carrier rejections, tells customers not to rotate caller IDs to get around reputation problems. And the number you swap in is not clean. Twilio's help center says a nuisance label is not removed when the previous owner releases a number, and Bandwidth says newly assigned numbers are mislabeled more often, especially on their first several calls, because they have no history at all.
So rotation buys you a fresh number with no track record, or a used one with someone else's, and either way the engines read the swap itself as evasion.
Spin up a pool of local presence numbers, spread the campaign across them, swap out the ones that get flagged.
Hiya grades number maturity, First Orion and Telnyx say do not rotate, and a swapped-in number carries its previous owner's label.
One aged number per purpose, two attempts a day at most, an agent that says its name in the first sentence.
Registered once, volume ramped instead of spiked, the same caller ID for the life of the campaign.
Who decides, and who does not
Three companies produce almost every label on a US mobile phone. First Orion is the analytics engine for T-Mobile, which shows Scam Likely and Potential Spam. Hiya is the engine for AT&T, which shows Spam Risk. TNS is the engine for Verizon, which shows Potential Spam. None of them is a carrier and none of them is a regulator, and the FCC's own staff wrote in a June 2021 call blocking report that the Commission does not regulate third-party analytics companies.
The rule everyone cites is 47 CFR 64.1200(k)(8). It requires a carrier that blocks calls to publish a single point of contact for blocking error complaints and to give a status update within 24 hours. Read the whole section, current to September 3, 2026, and the word label does not appear once. Paragraph (k) uses the word block, in one form or another, 34 times. That is not an oversight. In the December 2020 order that wrote the redress rule, the FCC said: "We decline to extend redress mechanisms to erroneous call labeling at this time." The commenters it credited for that outcome include CTIA, T-Mobile, First Orion and TNS. Two of the three companies behind the Free Caller Registry argued against giving you an appeal, and won. Twilio argued that a label is equivalent to a block, and lost.
Nothing has changed since. The FCC asked open questions about label accuracy in 2023 and adopted nothing. In May 2026 it asked, inside a proposed rule on caller ID attestation, whether calls signed at the top A level should be shielded from spam labeling. That is a question in a notice, not a rule, and reply comments close on September 8, 2026. Until something is adopted, the dispute routes are private: T-Mobile's call reporting form, where the support page says corrections follow within 48 hours of a verified request; AT&T's review-my-call-label page, which redirects to a Hiya ticket form; and Verizon's voice spam feedback site, which says a submission may not change how future calls are labeled.
What the Free Caller Registry actually does
The Free Caller Registry is one form that sends your business details and up to 20 numbers, or a spreadsheet of them, to all three engines at once. It costs nothing, and you should use it, after reading its terms. Registration does not guarantee redress, because each provider analyzes independently. Multiple submissions of the same number provide no additional value. And you must be the business that makes the calls, not a service provider, BPO or third party registering on a client's behalf, so an agency running AI calls for a brokerage either has the brokerage register or uses Hiya's separate on-behalf form.
The registry publishes no FAQ and no processing time, so the platforms fill the gap with their own numbers, and the numbers disagree. HighLevel's help center says the flag should be removed within four business days and that you can resubmit as many times as you want. Twilio's best practice article says the same process can take a few weeks. First Orion, which actually receives the submission, says removal has no fixed timeframe because determinations rest on trends across calls, not on a single request. Hiya's help center says it best: labeling is dynamic on every call, not a fixed property of the number. You are not appealing a verdict. You are feeding a score that recomputes on every dial.
Four moves before the next campaign
The bottom line
Spam Likely is a private score with no public appeal, set by three companies that asked the FCC not to referee them. The registry gets you identified. Aged numbers, capped attempts and an agent that says who it is in its first breath get you answered. That is how I build outbound AI callers, and the consent side of the same build is in my piece on whether AI cold calling is legal. Book a 30-minute call at cal.com/zeeshanwaheed/30min or email [email protected].
Frequently Asked Questions
Not by itself. It is one form that sends your business details and numbers to First Orion, Hiya and TNS, the three analytics companies behind the labels on T-Mobile, AT&T and Verizon. Its own terms say registration does not guarantee redress because each provider analyzes independently, and that submitting the same number more than once adds nothing. Register once, then fix the calling pattern the engines score.
No. Twilio and HighLevel both say authentication will not remove an existing nuisance label, Bandwidth says carrier analytics engines generally do not treat attestation as a primary factor, and Verizon says it is one factor and never the sole basis for a label. A-level signing on numbers you own raises the identity baseline. Call volume, duration, answer rates and complaints set the label.
No. Hiya grades a number on maturity, meaning whether you use established numbers without rotating, First Orion tells callers to avoid rotating or reactivating dormant numbers, and Telnyx tells customers not to rotate caller IDs to get around reputation problems. A newly bought number can also carry the previous owner's label, and Bandwidth says new numbers are mislabeled more often on their first several calls.
No. The FCC rule on call blocking, 47 CFR 64.1200(k)(8), requires a single point of contact and a 24-hour status update for blocking disputes and caller ID authentication effects, and the word label appears nowhere in the section. In December 2020 the FCC declined to extend redress to labeling, and a May 2026 proposal only asks whether A-attested calls should be shielded, with reply comments due September 8, 2026. Disputes go to T-Mobile's call reporting form, AT&T's review-my-call-label page, which is run by Hiya, and Verizon's voice spam feedback site.