The Voicemail That Doesn't Ring

Two county council races, one election-eve deadline, and 5,306 voters reached in their candidate's own voice. A case study and field guide to ringless voicemail: how it works, what the law requires, where it beats texting, and what to honestly expect.

A case study and field guide from Workforce Wave. Download the full 37-page PDF.

Two races, one day

On the Monday before Election Day, two county council candidates in the Southeast needed one more conversation with every supporter who had not yet voted. There was no time to knock doors, no budget for a phone bank, and voters had spent six weeks deleting political texts unread. The campaigns had one business day and a list of names.

The work came to Workforce Wave through one of the few political campaign partners we support. The partner runs strategy and voter data for local candidates; we run the delivery infrastructure underneath. For these two races the partner built the target universe, the candidates recorded their own messages, and our team turned the lists and the recordings into live ringless voicemail campaigns, handled the callback routing, ran a second delivery pass on every miss, and reported back the same night.

This document tells that story in detail and then steps back. Ringless voicemail is one of the least understood tools in modern outreach. It is often described with the same breathless confidence as every other channel, and it is often bought with the wrong expectations. We wanted to write down what it is, how it works at the carrier level, what it costs, what the law requires, where it beats texting and robocalls and where it does not, and what a campaign should realistically expect when it buys it. The two races are our evidence. The rest is what we have learned running the tool for candidates, nonprofits and businesses.

The assignment

Both candidates were running for seats on the same county council. Both were in contested races where turnout among their own likely supporters would decide the outcome. The partner's data team built each list from the county voter file using three filters.

Partisan lean. Voters scored as strong Republicans, lean Republicans, or genuine swing voters, the people most likely to support these particular candidates.
Vote history. Voters who had turned out in at least two of the last four general elections, so the list skewed toward people with a habit of voting who simply had not done it yet this cycle.
Not yet voted. Anyone who had already cast an early or absentee ballot was removed. Every dollar went to voters who could still be moved.

The final lists arrived the morning of the drop: 3,933 voters for the first race and 3,408 for the second, every record carrying a cell phone number. A handful of seed records, phones belonging to the partner and one candidate, were placed at the top of each list so the team could confirm with their own ears that the message was landing.

Each candidate recorded a short message in their own voice. One ran 30 seconds, the other 33. Both said, in plain words, who was calling, that tomorrow was Election Day, and why the vote mattered. Neither was produced in a studio. That turned out to be part of why they worked.

The clock

Morning	Final voter lists delivered. The delivery platform confirms the sending plan is active, 100 local sender numbers are attached, and both candidates' personal phones are verified as callback destinations.
Late morning	Lists cleaned, phone numbers normalized, duplicates removed, seeds moved to the front. Recordings transcribed by machine and checked word for word against the scripts. Every sender number mapped to exactly one candidate's callback line.
11:35 AM	Both campaigns launch, paced at 600 and 525 voicemails an hour, restricted to that Monday only so nothing could spill into Election Day.
1:00 PM	First check. About 45% of each list dialed, delivery tracking at the expected rate, the partner's seed phone has both messages.
5:34 PM	First pass complete. Every voter dialed. Delivery at 61.7% and 56.8%.
6:05 PM	Second pass launches against every voter whose first drop failed for a temporary reason, excluding anyone who already had the message.
8:09 PM	Second pass complete. Final delivery 74.3% and 69.9%. Reports built that night.

From list in hand to first voicemail delivered took under three hours. From launch to final numbers took under nine. That speed is the first thing worth understanding about this channel, and we will come back to why it is possible.

The results

5,306 voters received a personal voicemail from their candidate the day before the election
72.3% of both lists combined, after a second delivery attempt
87 voters called a candidate back the same day

	Race A	Race B	Combined
Voters on the list	3,933	3,408	7,341
Delivered on the first attempt	2,425 (61.7%)	1,935 (56.8%)	4,360 (59.4%)
Recovered on the second attempt	+499	+447	+946
Total delivered	2,924 (74.3%)	2,382 (69.9%)	5,306 (72.3%)
Voters who called back	40	47	87
Callback rate among voters reached	1 in 73	1 in 51	1 in 61

Three numbers in that table deserve more attention than they usually get.

The first-attempt rate was ordinary. Roughly six in ten voicemails landed on the first try. That is right where ringless voicemail lives on a clean cell-phone list, and anyone promising a campaign 85 or 90 percent on a single pass is either counting differently or hoping the campaign never checks. We explain where the other four in ten go in Part Two.

The second attempt was worth thirteen points. The evening retry recovered 946 voters who would otherwise have heard nothing. It cost nothing for the drops that failed again, because undelivered drops are not billed. It is the single cheapest improvement available in this channel, and many vendors do not offer it by default.

Eighty-seven people picked up their own phones and called a local candidate. On the day before an election, with no call to action beyond "call me back if you have questions," that is a meaningful signal. It is also where the campaigns left the most value on the table, which we cover in Part Five.

Who the message reached

Because every voicemail is tied to a record in the voter file, the results can be cut any way the file can. The clearest finding is how even the coverage was.

Segment	Race A delivered	Race B delivered
Strong Republican	74.6%	68.8%
Lean Republican	75.7%	69.3%
Swing	73.2%	71.4%
Voted in 4 of the last 4	76.1%	72.4%
Voted in 2 of the last 4	72.4%	69.0%
Age 18 to 34	75.7%	69.5%
Age 35 to 49	75.5%	73.6%
Age 50 to 64	74.3%	70.0%
Age 65 and older	72.0%	65.8%

Delivery barely moved across partisan segments or vote history. That is what you would expect from a technology that does not care who the voter is, only whether their carrier will accept a voicemail. Age is the exception. Older voters were the hardest to reach in both races, and in the second race the gap was nearly eight points. Older voters are more likely to have a voicemail box that was never set up or has filled up, and more likely to be on prepaid or legacy plans. They are also the voters most likely to read a mailer, which is one reason we argue in Part Eight that voicemail should never be a campaign's only contact.

Geography told a similar story. In the first race the best precinct reached 77.3% and the weakest large precinct reached 67.8%. In the second, the spread ran from 76.3% down to 62.7%, with one small rural precinct as low as 56.2%. Rural precincts with older populations and patchier carrier coverage trailed the newer suburban developments in both races.

Who called back

The callback data held the most interesting signal of the day. In each race, a different segment of the list was over-represented among the people who picked up the phone.

Segment	Share of Race A list	Share of Race A callers	Share of Race B list	Share of Race B callers
Strong Republican	43%	57%	43%	32%
Lean Republican	21%	10%	18%	21%
Swing	36%	32%	39%	47%

In the first race the candidate's base called back at a higher rate than anyone else. In the second race the swing voters did. A campaign cannot learn that from a text message open rate or from a robocall completion report. It is a small sample, 40 and 47 callers, and we would not stake a strategy on it. It does show how a voicemail program produces a second dataset, the people who chose to respond, that a campaign can use to decide where its candidate's time goes in the final hours.

One voter called the first candidate back twelve times in an afternoon. Most callers called once. Most calls were under fifteen seconds, which almost always means the voter reached the candidate's personal voicemail or hung up before anyone answered. Those 87 people wanted a conversation and, for the most part, did not get one in real time. We think that is the most important operational lesson of the two races.

How ringless voicemail actually works

A ringless voicemail is an audio message, either one recording for everyone or a message generated for each person, placed directly into a person's voicemail box without their phone ringing. The voter sees a new voicemail notification, the same red badge or banner they see when they miss a call from their dentist, and listens when it suits them. There is no interruption, no hang-up, no "Spam Likely" screen.

That description is simple. The plumbing underneath is not, and understanding it explains almost every result in Part One.

The voicemail box is a separate computer

When someone leaves you a voicemail in the ordinary way, three systems are involved: the caller's phone network, your carrier's switching network, and your carrier's voicemail platform. The voicemail platform is a server, usually in a carrier data center, that stores audio files for every subscriber and notifies their handset when a new one arrives. Your phone does not hold your voicemail. It holds a notification and a way to fetch the file.

Under normal conditions you only reach that server by calling the subscriber and letting the call fail over: they do not answer, the network gives up after a set number of rings, and the call is handed to the voicemail platform, which plays the greeting and records whatever you say.

Ringless voicemail skips the ringing step and talks to the voicemail platform directly. Providers have used a few different techniques over the years. The earliest placed two calls almost at once so the first kept the line busy and the second fell through to voicemail. Modern platforms generally connect to the carrier's voicemail system the way the network itself does when it forwards an unanswered call, delivering the audio as a deposit addressed to the subscriber's mailbox. The handset never sees an inbound call, so it never rings.

Three consequences follow from that design, and each one shows up in the data.

It only works on mobile numbers. Landlines, especially those on legacy copper or on business phone systems, use voicemail systems that ringless platforms cannot address the same way. A good platform checks each number before it dials and skips the landlines. In our two races, 17 numbers were skipped for this reason even after the list had been filtered for cell phones.
The voicemail box has to exist and have room. If a subscriber never set up voicemail, there is nothing to deposit into. If the box is full, the deposit is refused. Across both races 391 voters had no voicemail box at all and 325 had a full one. No amount of retrying fixes the first group.
The carrier decides. Every deposit is a request to a system owned by AT&T, Verizon, T-Mobile or one of the regional and prepaid carriers that ride on their networks. Carriers throttle, filter and occasionally refuse traffic for reasons of their own: volume from a given sender, the reputation of the sending number, network load at that moment, or policy. Those refusals come back as a generic error.

What happens to every voicemail

Every drop ends in one of a small number of outcomes. These are the categories we track and the ones we report back to every client.

Outcome	What it means	Retry?
Delivered	The carrier accepted the deposit and the voicemail is in the voter's box.	No. Never send twice.
Carrier error	The carrier refused or dropped the request without saying why. Usually temporary.	Yes. The best recovery category.
No pickup	The voicemail platform did not answer the deposit attempt in time.	Yes, with modest results.
Not detected	The connection completed but the platform could not confirm a voicemail greeting was reached.	Yes, with modest results.
Mailbox full	The box exists but has no room.	Sometimes. Boxes rarely clear within hours.
Mailbox not set up	The subscriber never activated voicemail.	No. Use another channel.
Not a cell phone	Landline or unsupported line type, caught before dialing.	No. Use another channel.
Number not in service	Disconnected or reassigned number.	No. Flag for list hygiene.

Here is how the first attempt broke down across both races, and what the second attempt did with each category.

First-attempt outcome	Voters	Retried	Delivered on retry
Delivered	4,360	0	n/a
Carrier error	1,881	1,881	about 45%
No pickup	256	256	about 10%
Not detected	245	245	about 23%
Mailbox full	273	273	under 10%
Busy or phone menu answered	16	16	too few to measure
Mailbox not set up	290	0	n/a
Not a cell phone, not in service	21	0	n/a

Retry recovery rates are measured on the early portion of the second pass, where outcomes were broken out by original failure reason. The final second-pass totals (946 recovered of 2,671 retried, 35.4%) are consistent with them.

Carrier errors are the story. They were the largest failure category on the first pass, about one in four dials, and nearly half of them went through a few hours later from the same sender pool. That pattern, a high error rate that drops sharply on a second try, is what you see when carriers are rate-limiting a burst of traffic rather than rejecting the numbers themselves. It is also why pacing matters. Sending the same 3,900 voicemails over eight hours instead of two would likely have produced fewer first-pass errors, and the retry is the cheapest way to recover the ones a faster pace produces.

The caller ID is a local phone number

Every ringless voicemail arrives from a phone number, and the voter sees it the same way they would see a missed call. That number matters more than most campaigns realize.

For these races we used a pool of 100 sender numbers, local to the county's area codes where possible, split between the two campaigns. The platform matched each voter to a sender number in their own area code when it could. People are far more likely to listen to a voicemail from a local number, and far more likely to call it back.

Spreading traffic across many numbers also protects deliverability. Carriers watch how much traffic each number generates. A single number sending thousands of deposits in an afternoon looks like what it is. One hundred numbers sending about 75 each looks like a busy local office. The platform's guidance was roughly 50 sends per number per day as a comfortable ceiling. We ran at about 75 on a one-day burst, accepted somewhat higher error rates in exchange for speed, and recovered most of the difference on the retry.

Each sender number in this program forwarded calls to exactly one candidate's personal phone. When a voter called back, the call rang the right candidate, not a shared campaign line and not a stranger. Before launch we checked every one of the 100 numbers against its routing profile and confirmed that none pointed anywhere else. That check takes twenty minutes and prevents the worst possible outcome of a political voicemail program, a voter returning one candidate's call and reaching another.

Pacing, windows and the Election Day wall

Ringless platforms let a campaign control three things about timing: which days messages can go out, the hours of the sending window, and how many go out per hour. For a GOTV push the day before an election, the days setting is the most important safety control in the system. Both campaigns were restricted to that Monday only. Had anything stalled, the platform would have held the remainder rather than sending "tomorrow is Election Day" on Election Day.

The sending window was 11 AM to 7 PM in each recipient's own time zone, later extended to 8 PM for the retry. Hourly rates were set so each list would finish inside the window with room to spare, then raised when the launch slipped past 11. One subtlety worth knowing is that most platforms treat the end of the window as a soft stop. They stop queuing new drops at the cutoff but let drops already in flight finish, which can run 20 to 30 minutes past the stated end. A campaign that needs a hard stop at a certain time should set the window to end half an hour earlier.

Why it can launch in hours

A direct mail piece takes a week from approved proof to mailbox. A door-knocking program needs volunteers, turf maps and a weekend. A peer-to-peer texting program needs a registered brand and campaign with the carriers, which can take days to weeks the first time. Ringless voicemail needs a clean list of cell numbers, an audio file, a set of sender numbers and a destination for callbacks. If those four things exist, a campaign can go from list to first delivered voicemail in the time it takes to upload a CSV and click through a setup screen. On election eve, that is the feature that matters.

Voicemail against every other channel

No campaign chooses a channel in isolation. Whether ringless voicemail "works" matters less than what it does better or worse than the other things the same budget could buy. This part compares it with each of the alternatives a local campaign actually weighs in the last weeks of a race.

The phone has stopped ringing for strangers

Every phone-based channel is fighting the same behavior. In a 2020 Pew Research Center survey, about eight in ten American adults said they generally do not answer their cell phone when an unknown number calls.⁶ That single habit is why live phone banks and robocalls have lost so much of their reach over the past decade, and it shapes everything else in this part.

A second Pew finding from the same year matters even more for this document. When asked about their general practice, 67 percent of Americans said they do not answer calls from unknown numbers but do check the voicemail if one is left.⁷ That is the opening ringless voicemail is built for. The voter screens the call that never actually rang, and then listens to the message on their own time.

Texting

Political texting exploded after 2018, and it is now the default contact channel for many campaigns. It has real strengths. Nearly everyone reads at least the preview of a text. A text can carry a link to a polling place lookup, a donation page or a volunteer form. A reply comes back in the same thread.

Its weaknesses have grown just as fast.

Volume and fatigue. Americans received an estimated 15 billion political texts in 2022, a record at the time.⁸ Call-blocking firms estimated billions more in 2024, with monthly volume in September alone several times higher than early summer.⁹ In the final weeks of a cycle, a political text is competing with dozens of others, many of them asking for money.
Registration and filtering. Application-to-person texting from ordinary ten-digit numbers runs on a carrier registration system known as 10DLC. Political senders typically need a vetting token from Campaign Verify, a nonprofit that checks public campaign filings, before their brand and messaging campaign can be registered with The Campaign Registry.⁵ Unregistered or poorly registered traffic is filtered by carriers, sometimes silently. A first-time campaign that starts the process late can find itself unable to text at all in the final week.
The consent line. Federal rules treat automated texts as calls. Texts sent by a human pressing send for each message are generally treated differently, which is why political texting is dominated by "peer-to-peer" platforms where volunteers click through messages one at a time.³⁴
The words are not the candidate. A text is a text. It can be signed with the candidate's name, but it does not carry their voice, and voters know the candidate did not type it.

The research on texts as a turnout tool is encouraging when volume is low. A field experiment in the 2006 election found that a text reminder the day before raised turnout by about three percentage points among the voters who received it.¹⁰ That experiment ran when a political text was still a novelty. Whether a GOTV text in 2026, arriving alongside a dozen others, carries the same effect is an open question that campaigns should answer with their own tests rather than assume.

Robocalls

A robocall is a prerecorded message delivered by calling the voter's phone and playing the audio when someone, or an answering machine, picks up. Ringless voicemail shares its economics and much of its legal treatment, but the experience is very different.

A robocall interrupts. It rings, often while the voter is driving or working, and frequently shows up with a carrier spam warning. The voter answers, realizes it is a recording, and hangs up within a few seconds, or never answers at all. Political robocalls to residential landlines are permitted without prior consent within limits, but calls to cell phones generally require the called party's prior express consent.³ In a 2020s electorate where most voters are reachable only by cell, that leaves robocalls with a shrinking audience of landline households, which skew older.

The academic evidence on robocalls as a turnout tool has never been strong. The most widely cited body of field experiments, summarized in Donald Green and Alan Gerber's Get Out the Vote, has found little or no measurable turnout effect from conventional robocalls, in contrast to door-to-door canvassing and live volunteer calls.¹¹ A ringless voicemail is not a robocall in the voter's experience, since it never interrupts them, but it is honest to say that no published field experiment we know of has isolated ringless voicemail's effect on turnout. That is a gap in the evidence, and one we would like to help close.

Live calls and phone banks

A live conversation is far more persuasive than any recording. Field experiments consistently find that volunteer phone calls, particularly from people with some connection to the voter, move turnout more than any automated contact.¹¹ The constraint is reach. If eight in ten voters do not answer unknown numbers, a phone bank spends most of its time listening to ringing and leaving voicemails by hand. A volunteer might complete eight or ten real conversations an hour.

The interesting combination is a ringless voicemail first and a live call to the people who call back. The voicemail does the reaching. The callback does the conversation, initiated by the voter, at a moment they chose.

Door to door

Canvassing is the gold standard. In Green and Gerber's summary of decades of experiments, a well-run volunteer canvass produces roughly one additional vote for every 14 people actually contacted at the door.¹¹ Nothing automated comes close. Canvassing is also slow, weather-dependent, volunteer-intensive and nearly impossible to execute across a whole district in the final 24 hours. Most local campaigns use it for their highest-priority precincts and need something else for everyone else.

Mail

Direct mail reaches every household with an address, cannot be filtered by a carrier and is the channel older voters trust most. It also costs far more per piece than any phone channel, takes days to produce and deliver, and lands in a mailbox crowded with other campaigns' pieces in the final week. A mail piece that arrives the day after the election does nothing. For an election-eve message, mail has to be planned and dropped a week or more in advance, using a list that will be out of date by the time it lands.

Side by side

	Ringless voicemail	Peer-to-peer text	Robocall	Live call	Door	Mail
Candidate's own voice	Yes	No	Yes	No (volunteer)	Sometimes	No
Interrupts the voter	No	Lightly	Yes	Yes	Yes	No
Typical reach of a cell list	60 to 75%	High, minus filtering	Low on cells	Low (answer rates)	Limited by volunteers	Nearly all addresses
Setup time	Hours	Days to weeks first time (registration)	Hours	Days (volunteers)	Weeks	A week or more
Can carry a link	No	Yes	No	No	Literature	Printed URL or QR
Two-way	Via callback	Via reply	Rarely	Yes	Yes	No
Proof of attention	None (delivery only)	None (delivery only)	Partial (answered)	Yes	Yes	None
Relative cost per contact	Low	Low	Lowest	High	Highest	High
Evidence on turnout	Untested in published research	Positive in early studies	Weak	Positive	Strong	Modest

Read across that table and a pattern appears. Ringless voicemail is the only channel that combines the candidate's voice, no interruption, same-day setup and low cost per contact. It gives up proof of attention and the ability to carry a link. Text gives up the voice and gains the link. That is the core of the argument in Part Eight for using the two together, with voicemail first and text picking up the voters voicemail could not reach.

From one recording to one voter: the personalization ladder

The two races in this study used one recording per candidate, the same 30 seconds for every voter. That was the right call for an election-eve reminder on a tight clock. But a ringless voicemail no longer has to be the same message for everyone. The same platform can build each message on the fly, dropping in the voter's first name, their neighborhood, their polling place or the issue they care about, so the voicemail sounds like the candidate is speaking to that one person.

That is the most interesting thing about this channel right now, and it is where we expect most campaigns to go next. Here is the full ladder, from the simplest message to the most personal.

Rung one: the candidate's own recording

The candidate records a message on a phone or a decent microphone. The file is uploaded once and delivered unchanged to every voter.

The advantages are larger than they look. It is the candidate's real voice, with real cadence and the small imperfections that make a message feel like a call from a neighbor rather than an ad. It can be recorded in five minutes. It costs the least to deliver, because platforms meter static audio at a fraction of the cost of generated audio. There is nothing to go wrong in production. And there is no regulatory ambiguity about whether the voice is artificial, because it is not.

The limitation is that every voter hears the same words. The message cannot say "Hi Maria" or "your polling place is the library on Main Street." For a GOTV reminder from a local candidate, that limitation rarely matters. Both candidates in this study recorded their own messages, and the callback rates suggest voters heard a person.

Recording tips that measurably help. Say your name and the office in the first five seconds, because many voters decide in that window whether to keep listening. Speak at a normal conversational pace, slightly slower than feels natural. Keep it under 35 seconds. Name one specific action ("vote tomorrow, polls are open seven to seven"). Say the callback number slowly, or say "call me back at this number," since the caller ID already shows it. Record somewhere quiet with soft surfaces, such as a parked car or a closet, rather than a bare office.

Rung two: segmented static recordings

The candidate records several versions, and each segment of the list gets the version written for it. A swing-voter version talks about one local issue. A base version talks about turnout. An early-vote version mentions the early voting site nearest that part of the district. A Spanish-language version goes to voters flagged for it in the file.

This is still static audio, so it carries the same low cost and zero production risk as rung one. The only added work is writing and recording three or four scripts and splitting the list. For many local campaigns, this is the best value on the ladder.

Rung three: dynamic messages built for each voter

Here the message stops being a recording and becomes a script with blanks in it. "Hi {First Name}, it's the campaign for {Race}. I know {Issue} matters in {Neighborhood}. Your polling place tomorrow is {Polling Place}, and polls are open seven to seven." For every voter, the platform fills in the blanks from their record and generates the audio before delivering it. Any column in the list can become a blank: first name, street or neighborhood, precinct, polling place, early-vote status, the issue a canvasser logged at the door.

The difference in how it lands is large. A generic reminder sounds like an ad that happened to arrive as a voicemail. A message that opens with the voter's own name sounds like someone who knows them took the time to call. People listen past the first five seconds when they hear their name, and a voter who feels spoken to is far more likely to call back.

The economics are better than most people assume. Generated audio is metered by length, so a dynamic drop costs more than a static one, but on a well-written script of about 450 characters it runs roughly twice the per-drop cost of a recording, not an order of magnitude more. Platforms that compress the billable length of generated scripts help keep it there; that should be checked against the actual script before the campaign is priced.

The quality of the data now matters to the sound of the message. A first name stored as "WILLIAM J" in the voter file will be read aloud as "William J." A polling place that changed since the file was pulled will be wrong in the voter's ear. Merge-field programs need a data cleanup pass that static programs do not.

Rung four: the candidate's cloned voice

The top of the ladder combines the warmth of rung one with the personalization of rung three. The candidate records a sample, typically a few minutes of clean speech, and a voice model is trained to sound like them. Every personalized message is then generated in that voice. A voter hears the candidate say their name.

Voice cloning has become remarkably good, and on a phone-quality voicemail most listeners cannot tell a well-made clone from a recording. That is precisely why it carries the most obligations. Federal regulators treat AI-generated voices in calls as artificial voices under the same law that governs robocalls, which brings consent and disclosure requirements we cover in Part Six. Several states have passed or proposed laws requiring disclosure when synthetic media is used in political communication. A campaign that uses a cloned voice should disclose it in the message, should have the candidate's written authorization on file, and should have counsel review the script.

Used carefully, a cloned voice is a powerful tool for a candidate who cannot personally record thousands of variations. Used carelessly, it is a news story. We recommend it for campaigns that have counsel, a clear disclosure line in the script and a reason to personalize that a segmented static recording cannot meet.

Choosing a rung

	Own recording	Segmented recordings	Dynamic, synthetic voice	Dynamic, cloned voice
Sounds like	The candidate	The candidate	A natural AI narrator	The candidate
Says the voter's name	No	No	Yes	Yes
Relative cost per drop	Lowest	Lowest	About double	About double
Setup time	Minutes	An hour or two	Hours, plus data cleanup	A day or more
Compliance load	Standard	Standard	Higher: artificial voice	Highest: artificial voice plus likeness
Best for	Election eve, small budgets	Persuasion, early vote, multilingual	Polling place and logistics reminders	High-stakes personal outreach with counsel

Our rule of thumb: if the list has clean first names, use them. A static recording is the fastest option when the clock is short, as it was on election eve here. When there is a day or two to prepare, a message that greets each voter by name and points them to their own polling place is the better product, and the cost difference is small next to the lift in attention and callbacks.

Interactive messages and AI on the other end

Personalization is not only about the outbound message. The more significant development is what happens when the voter calls back. A ringless voicemail can end with "call me back at this number," and that number can be answered by a person, by voicemail, or by an AI voice agent trained on the campaign's positions, polling locations, hours and volunteer sign-ups. That is the subject of the next part, and in our view it is where most of the unrealized value in this channel sits.

The callback is the hidden channel

Every other form of automated outreach is one-directional. A mailer cannot be answered. A broadcast text can be replied to, but replies to political texts are dominated by STOP requests and arguments. A robocall can be hung up on. A ringless voicemail arrives from a real phone number, in what sounds like a personal message, and some fraction of voters do the most natural thing in the world and call it back.

In our two races that fraction was 1 in 73 voters reached for the first candidate and 1 in 51 for the second. Eighty-seven people in total. They were not random. They skewed toward the candidate's base in one race and toward swing voters in the other, and they included people who called more than once.

What happened when they called

Every callback was routed to the right candidate's personal phone. The routing worked perfectly; we verified each of the 100 sender numbers before launch, and the platform logged every connected call. What the routing could not do was answer the phone.

Callback detail	Race A	Race B
Voters who called back	40	47
Total calls	56	51
Voters who called more than once	6	4
Combined talk time	6 min 32 sec	9 min 44 sec
Longest single call	27 sec	36 sec
Calls of 15 seconds or more	5	13

The average call lasted about nine seconds. That is long enough to hear a voicemail greeting and hang up. A candidate on the day before an election is driving between events, knocking doors and doing interviews. They cannot answer forty calls in an afternoon, and they should not try. The likely result is that most of the voters who reached out heard a voicemail greeting, and some of them called back again and again. One voter called the first candidate twelve times.

None of this is a failure of the voicemail program. It is the program working well enough to create a demand the campaign was not set up to meet.

Three ways to answer the callback

Route to a staffed line. A campaign office or a volunteer with a cell phone answers. This is the most personal option and the hardest to staff, since callbacks arrive in bursts that follow the sending pace. A volunteer can handle a few callbacks an hour well. They cannot handle the twenty that arrive in the first hour after a large drop.

Route to the candidate, with a plan. This is what both campaigns did, and it is reasonable for a small race, provided someone checks the candidate's voicemail every hour and returns calls. We provided both candidates with a list of every voter who called, with phone numbers and precincts, sorted so repeat callers came first. A returned call from the candidate the night before the election is about as strong a contact as a campaign can make.

Route to an AI voice agent. The callback number is answered by a voice agent that knows the candidate's platform, the polling locations and hours, how to request a ride to the polls, and how to sign up to volunteer. It answers every call immediately, at any volume, at any hour. It can take a message for the candidate, flag the callers who want a personal conversation, and send a text with the polling place address. When a voter asks something it should not answer, it says so and takes a callback number.

We build and run AI voice agents for businesses, municipalities and nonprofits, and we believe the third option is where political voicemail is heading. Inbound calls that a voter initiates sit in a very different regulatory position from outbound automated calls, because the voter chose to call. The agent can be transparent that it is an AI assistant for the campaign. And it turns a ten-second hang-up into a real exchange of information, every time.

What we would change next time. For any program larger than a few hundred voters, we would put an AI agent or a staffed line in front of the candidate's phone, with an option to transfer to the candidate when it matters. We would also add one sentence to the voicemail script telling voters what happens when they call ("call back and my team will help you find your polling place"), because a clear reason to call is the strongest lever on callback rate.

Callbacks as a measurement tool

The deepest weakness of ringless voicemail is that nobody can prove a voter listened. Platforms report deliveries, not plays, because carriers do not report when a voicemail is opened. Callbacks are the one direct behavioral signal the channel produces. They tell a campaign which message, which segment and which precinct prompted people to act.

Over several drops, callback rate becomes a useful comparison metric. If the swing-voter script produces twice the callbacks of the base script, that tells the campaign something about which message is landing. If one precinct calls back at three times the district average, that is where the candidate should spend the last afternoon. In our races, the segment that over-indexed on callbacks differed between the two candidates. With only one drop each, that is an observation, not a finding. Across a cycle with three or four drops, it would be a strategy input.

The rules

Ringless voicemail spent years in a gray zone, sold on the theory that a message that never rings a phone is not a phone call. That theory is gone. Any campaign buying the channel today should understand the federal framework, the state overlay and the specific obligations that come with AI-generated voices.

A note before the details. What follows is a practitioner's summary, not legal advice, of the rules as we understand them in late 2026, written to help campaigns ask their counsel the right questions. Federal and state rules change, and enforcement turns on facts. Every campaign should have counsel review its list, its consent basis and its scripts before a drop.

A ringless voicemail is a call

In November 2022 the Federal Communications Commission issued a declaratory ruling, FCC 22-85, holding that ringless voicemail delivered to a wireless phone is a "call" under the Telephone Consumer Protection Act, and specifically a call made using an artificial or prerecorded voice.¹ The ruling applies to any provider that uses a subscriber's mobile number to direct a message to the voicemail box associated with that number. The practical meaning is simple. Every rule that governs a prerecorded call to a cell phone governs a ringless voicemail to a cell phone.

Prerecorded political calls to cell phones

Political calls get some latitude under federal law that commercial calls do not. They are not "telephone solicitations," so they are generally exempt from the National Do Not Call Registry, and prerecorded political calls to residential landlines are permitted without prior consent within frequency limits.³ Calls to cell phones are treated differently. The FCC's guidance to political campaigns is that autodialed or prerecorded calls and texts to wireless numbers require the called party's prior express consent, and that a caller claiming consent carries the burden of proving it.³

What counts as prior express consent in a political context is the question every campaign should put to counsel. Consent can come from a voter providing their cell number to the campaign or party, from sign-up forms with appropriate language, and from other relationships recognized in FCC guidance and case law. Voter-file phone numbers appended from commercial data sources are the area of greatest risk. A campaign should know where every number on its list came from and on what basis it is contacting it.

Identification and callback number

Prerecorded messages must identify who is responsible for the call and must provide a telephone number for the calling party, during or after the message.³ For a campaign that means stating the name of the candidate or committee at the beginning of the message and making sure a working callback number is available. In a ringless voicemail program the sender number itself serves as the callback number, which is one more reason that number must actually route somewhere a person or an agent will answer.

Time of day

Federal "quiet hours" (no calls before 8 AM or after 9 PM in the recipient's local time) are written into the rules for telephone solicitation, which political calls are not. Responsible political callers follow them anyway, and a number of states impose their own time restrictions that do apply to political calls. Our standing practice is to schedule every drop within the recipient's local daytime hours, using the platform's recipient-time-zone setting, and to end the window well before evening. In the case study, sending ran from 11 AM to 8 PM recipient time.

AI-generated and cloned voices

In February 2024, weeks after an AI-generated voice imitating a national political figure was used in robocalls to voters ahead of a primary, the FCC unanimously ruled that AI-generated voices are "artificial" voices under the TCPA.² The ruling did not create a new category of illegal call. It confirmed that a call using an AI voice is subject to every rule that applies to artificial or prerecorded voice calls: consent, identification, callback number and the rest. Combined with the 2022 ringless voicemail ruling, the result is that an AI-voiced ringless voicemail to a cell phone sits squarely inside the TCPA.

States have moved further. By early 2026 roughly 28 states had enacted laws addressing AI-generated or "deepfake" content in political communications, many of them requiring a disclosure when synthetic media depicts a candidate.¹³ These laws vary widely in what they cover, how close to an election they apply and what the disclosure must say. A campaign using a cloned version of its own candidate's voice should assume a disclosure is required, should get the candidate's written authorization for the clone, and should have counsel check the specific statute in its state.

Texting has its own rules

Because Part Eight recommends texting as the follow-up channel, it is worth summarizing its rules too. Automated texts are treated as calls under the TCPA. In 2021 the Supreme Court narrowed the definition of an "automatic telephone dialing system" in Facebook v. Duguid to equipment that uses a random or sequential number generator,⁴ which reshaped the legal analysis around texting platforms. The FCC's guidance to political campaigns remains that texts sent manually by a person do not require prior consent while automated ones do.³ On top of the law sits the carrier registration system described in Part Three, which is not a legal requirement but determines whether messages are delivered at all.

A compliance checklist for a voicemail drop

Document the source of every phone number on the list and the consent basis for contacting it.
Identify the candidate or committee by name at the start of the message.
Make sure the callback number works and is answered or reaches a voicemail box that is checked.
Honor opt-out requests from callbacks, texts or any other channel, and suppress those voters from future drops.
Schedule in the recipient's local time, within daytime hours, and check state-specific rules.
For any AI-generated or cloned voice, include a clear disclosure, keep the candidate's written authorization on file, and have counsel review against state synthetic-media law.
Keep records of what was sent, to whom and when, for at least as long as counsel recommends.

Where voicemail fits in a campaign calendar

The two drops in this study were election-eve GOTV, the most common use of ringless voicemail in politics and the one where its speed matters most. It is far from the only use. A campaign that plans for voicemail from the start can use it four or five times in a cycle, each with a different job.

Introduction, months out

A first-time local candidate has a name recognition problem. A 30-second voicemail in the candidate's own voice, saying who they are, why they are running and one thing they care about, does something a yard sign cannot. It puts a voice to a name. Early in the cycle, when voters are not yet saturated with political contact, a voicemail from an unfamiliar local number is more likely to be heard out. This is also the drop that builds the callback list: voters who respond to an introduction are prime volunteer and donor prospects.

Persuasion, six to three weeks out

Segmented static recordings (rung two in Part Four) earn their keep here. Swing voters hear about the issue the campaign's research says moves them. Base voters hear about stakes and turnout. A message about a specific local issue, such as a road project or a school rezoning, delivered to the precincts it affects, is about as targeted as a campaign can get without a volunteer on the porch.

Persuasion is also where honesty about the channel matters most. Voicemail is good at delivering a message to an identified voter. It is weaker at changing minds than a conversation, because it is one-way. We would never sell a voicemail drop as a substitute for persuasion canvassing. We would sell it as the way to reach the persuadable voters the canvassers will not get to.

Early vote and absentee chase

In states with early voting, the window opens with a logistics problem: voters do not know where early voting sites are or when they are open. A merge-field voicemail (rung three) that names the early voting site nearest the voter and its hours is a genuinely useful service, the kind of message voters thank campaigns for. For absentee chase, a voicemail to voters who requested a ballot but have not returned it, reminding them of the deadline and how to return it, is one of the highest-value uses in the entire cycle. Each week the campaign should pull an updated voter file, remove those who have voted, and drop again on the remainder.

GOTV, the final 72 hours

This is the scenario in our case study. The universe is supporters and leaners who have not yet voted. The message is short and specific: who is calling, that the election is tomorrow (or today, for an Election Day morning drop where state rules and the campaign's own judgment allow it), the polling hours, and an invitation to call back. Speed is everything. A same-day build lets the campaign use the freshest possible list, removing everyone who voted early right up to the morning of the drop.

For these drops we strongly recommend the day restriction we used in our races. The platform should be told explicitly which day or days the drop may run. A message that says "tomorrow is Election Day" must not be delivered on Election Day, and a stalled queue must not resume the next morning.

After the election

Winning candidates have a reason to say thank you, and a voicemail from a newly elected official thanking the voters who supported them costs little and builds a relationship for the next cycle. Campaigns that lose can use the same channel to thank volunteers and supporters. Either way the callback list from the cycle becomes the foundation of the next one.

Phase	Typical universe	Message	Best rung
Introduction	Likely voters in the district	Who I am and why I'm running	Own recording
Persuasion	Swing and soft-support segments	One issue, one segment	Segmented recordings
Early vote	Supporters who have not voted	Where and when to vote early	Dynamic messages
Absentee chase	Requested but not returned	Deadline and how to return	Dynamic messages or own recording
GOTV	Supporters and leaners not yet voted	Vote tomorrow, hours, call me	Own recording
Thank you	Supporters, volunteers, callers	Thank you, stay in touch	Own recording

Full coverage: what to do about the voters voicemail cannot reach

After two attempts, 2,035 of the 7,341 voters in our study had still not received a voicemail. That is 28 percent of two carefully targeted lists of people the campaigns badly wanted to reach. Every ringless voicemail program leaves a remainder like this. The question is what the campaign does about it.

The answer starts with the outcome codes. A voicemail program does not just fail to reach people; it tells you why, voter by voter, and the reason points straight to the next channel.

Sorting the remainder

Why the voter was not reached	Voters	What it tells you	Next channel
Carrier error after two attempts	995	The number is probably fine; the carrier refused this traffic.	Peer-to-peer text, or a third voicemail attempt on a slower pace
No voicemail box	391	The phone is active but cannot take voicemail.	Text first. These voters read texts, since they have no voicemail.
Mailbox full	325	The voter does not clear voicemail. They may not listen to it either.	Text, then mail or door
No pickup or not detected	299	Inconclusive; often a carrier or voicemail platform quirk.	Text
Not a cell, not in service	25	The phone number on file is wrong or a landline.	Mail, door, or a live call to a landline

Counts are final outcomes after both attempts, combined across both races. Busy and phone-menu outcomes are included with no pickup.

Almost every unreached voter in our study still had a working cell phone. That makes texting the natural follow-up, and the combination works better than either channel alone. A voter who got the voicemail does not need the text. A voter whose voicemail failed gets the text instead, and the campaign does not double-contact anyone. The two channels share the same list, the same targeting and the same phone numbers, so the handoff is a filter, not a new data project.

A layered contact plan

Here is the structure we recommend for a GOTV push that aims to touch every voter on the list once, through the most personal channel that will actually reach them.

Ringless voicemail to the full list. The candidate's voice, a clear ask, a callback number. Delivered to roughly six in ten voters on the first pass.
Same-day voicemail retry on carrier errors and inconclusive outcomes, at a gentler pace. Adds ten to fifteen points of coverage at no cost for repeat failures.
Peer-to-peer text to every voter not delivered after the retry, with a link to the polling place lookup and the same callback number. Sent by a person clicking send for each message, which is how most campaigns keep political texting outside the federal autodialer rules (see Part Six).
Mail or doors for voters with no working cell number, and for the oldest voters, where both voicemail and text perform worst and mail performs best.
A live answer on the callback line, by staff, volunteers or an AI agent, so every voter who responds to any of the above reaches someone.

Layering works because the channels fail in different places. Voicemail fails on full and missing mailboxes, and those voters still read texts. Text fails on carrier filtering and on voters who ignore political texts, and those voters often listen to a voicemail from a local number. Both fail on wrong numbers, and mail and doors do not depend on phone numbers at all.

Supplemental strategies worth knowing

Matched digital ads. The same voter list can usually be uploaded to the major ad platforms as a custom audience. Match rates vary, often reaching only about half of a list, but a short video of the candidate saying the same thing as the voicemail, shown to the same voters in the last 48 hours, reinforces the message for the voters the voicemail reached and provides a second chance at those it did not.

Relational outreach. Volunteers text or call people they personally know who are on the list. It is the most persuasive contact a campaign can make and the hardest to scale. The voicemail callback list and the list of voters who could not be reached are both natural assignments for relational volunteers.

Live calls to the callbacks. The 87 voters who called back in our study self-identified as engaged. A volunteer who calls each of them back personally, even after the election, is converting them into volunteers, donors and repeat voters.

A third voicemail attempt on a different day. Where the calendar allows, a third attempt one or two days later, slower and spread across more sender numbers, recovers more of the carrier-error group. In a GOTV window there is rarely time. In a persuasion or early-vote window there almost always is.

Interactive voicemail. The callback itself can be built into a campaign tool. "Press nothing, just call this number back, and we will tell you your polling place." The voicemail becomes the doorway into an AI-answered service line that can look up polling places, schedule a ride, take a volunteer sign-up or record a question for the candidate.

The test of a contact plan. For every voter on the list, the campaign should be able to say which channel reached them, or why none did. A voicemail program that reports outcome codes voter by voter makes that possible. A program that only reports a delivery percentage does not.

Beyond politics

Political campaigns are the loudest users of ringless voicemail, but they are not the only ones, and in some ways they are the least typical. A campaign wants to reach thousands of people once, on a hard deadline. Most other organizations want to reach the same people repeatedly, with information those people actually asked for. That changes the economics, the consent picture and the results.

Healthcare

Missed appointments are one of the most expensive problems in outpatient care, and reminders are one of the best-studied fixes. A widely cited systematic review of telephone and text reminders for hospital appointments found that reminders reduced non-attendance by about a third relative to baseline, with automated reminders reducing it by 29 percent and manual phone calls by 39 percent.¹² Voicemail suits this job well. A reminder in the practice's own voice, left without interrupting the patient at work, with a callback number that reaches scheduling, is a natural fit. Healthcare also carries its own privacy rules on what a message may say, which favor short, generic messages ("this is a reminder from Dr. Lee's office about your appointment tomorrow; please call us at...") over anything that discloses a condition.

Local government and utilities

Cities and counties need to reach residents about things residents care about: a boil-water notice, a trash pickup change, a road closure, a property tax deadline, an evacuation route. Many residents never sign up for alert apps. A voicemail from the city's own number, in the voice of a known official, reaches them on the device they already carry. Paired with an AI agent on the callback line that can answer follow-up questions around the clock, it turns a one-way announcement into a service. We run AI voice agents for municipalities that handle exactly these callback questions.

Nonprofits and associations

A nonprofit asking for event attendance, membership renewal or volunteer shifts faces the same text fatigue and unanswered calls as a campaign, with a fraction of the budget. A voicemail in the executive director's voice thanking a donor, or reminding a volunteer about Saturday's shift, is personal in a way an email blast is not. Because these are typically messages to people with an existing relationship with the organization, the consent picture is often cleaner than in politics.

Home services and small business

Plumbers, HVAC companies, painters and other service businesses use voicemail to reach past customers with seasonal reminders (furnace check before winter, gutter cleaning in the fall) and to follow up on estimates. The callback is the entire point. A homeowner who calls back is ready to book, and an AI agent or a dispatcher on the other end converts that call while the intent is fresh. Commercial voicemail is held to stricter consent rules than political voicemail, so these programs are built on customer lists with documented relationships.

What carries over from politics

Every lesson in this case study applies outside politics. Clean the list. Keep cell numbers only. Expect about six in ten on the first attempt and retry the temporary failures. Spread traffic across local sender numbers. Route every callback somewhere it will be answered. Report outcomes by reason so the remainder can be reached another way. The regulated nature of the channel applies everywhere too. Commercial senders face tighter consent requirements than campaigns, not looser ones.

Limitations and honest expectations

Every outreach channel is oversold, and ringless voicemail more than most, because its central promise (your message, in your voice, in every voter's pocket, without ringing their phone) sounds like it should work every time. It does not. Campaigns that understand the limits buy it for the right jobs and are pleased with the results. Campaigns that do not, end the cycle feeling cheated by a channel that did exactly what it does.

What to expect, in numbers

Measure	Realistic expectation	Our two races
Delivery, first attempt, clean cell list	55 to 68 percent	61.7% and 56.8%
Delivery after a same-day retry	68 to 78 percent	74.3% and 69.9%
Voters whose voicemail box is missing or full	8 to 12 percent of a cell list	9.8%
Callback rate among voters reached	0.5 to 3 percent, driven by the script and the candidate	1.4% and 2.0%
Time from final list to first delivery	Hours, once sender numbers and routing exist	Under three hours
Listen rate	Unknown. Carriers do not report plays.	Unknown

Expectation ranges reflect our experience across political and commercial ringless voicemail programs on filtered cell-phone lists. Lists with stale numbers, landlines or VoIP numbers will run lower.

The limits that matter most

Nobody can prove the voter listened. This is the fundamental limitation of the channel and it is permanent. A delivered voicemail is in the voter's box. Whether they played it, played the first five seconds, or deleted it from the notification without hearing a word is invisible to everyone except the voter. Any vendor that reports "listens" or "opens" for ringless voicemail is estimating. The two indirect signals are callbacks, which are real but small, and turnout among the voters reached compared with a held-out control group, which is the gold standard and which very few local campaigns run.

Delivery is not 100 percent and never will be. Between missing mailboxes, full mailboxes, carrier refusals and wrong numbers, roughly a quarter to a third of a good list will not receive a voicemail after two attempts. Plan the follow-up channel before the drop, not after.

Older voters are harder to reach. In our races the oldest age group trailed the best-performing group by four to eight points. Older voters are also the most reliable voters in most electorates, so this gap matters. Mail remains the strongest channel for them.

Speed costs some deliverability. Pushing thousands of voicemails through in a few hours raises carrier error rates. A retry recovers much of it, but a campaign that can spread a drop over two days will get a higher first-pass rate than one that compresses it into an afternoon.

The message is one-way. A voicemail cannot answer a question, counter an objection or hear a concern. It delivers a message. Persuasion happens in conversation, and the callback line is the only route to one.

Frequency fatigue is real. A voter who gets a voicemail from a campaign once listens. A voter who gets one every three days stops, and may start resenting it. We recommend no more than one ringless voicemail per voter per week from a single campaign, and fewer outside the final two weeks.

It is a regulated call. Federal regulators treat a ringless voicemail as a call. That brings consent rules, identification requirements, time-of-day limits and, for artificial voices, additional obligations. Part Six covers them. A vendor that tells a campaign voicemail drops are "not calls" is repeating a position regulators rejected in 2022.

Your list is your ceiling. Voicemail can only reach the numbers on the file. Voter files carry phone numbers appended from commercial sources, and those numbers go stale. In our study 25 numbers were dead or not cell phones even after filtering. On less carefully prepared lists we have seen that number run far higher.

What success looks like

A successful ringless voicemail program, for a campaign with realistic expectations, looks like this. Seven in ten targeted voters receive a short personal message from the candidate at the moment it matters. Every voter who could not be reached is identified by reason and handed to another channel. A small number of highly engaged voters call back and are answered. The campaign spends money only on messages that were actually delivered, and knows by the end of the day exactly who received what.

That is what happened in both of our races, with one exception. The callbacks were routed perfectly and answered rarely. Fix that, and you have the program we would run for any candidate.

The playbook

This is the checklist we run for every ringless voicemail program, political or commercial. Most of it is unglamorous. Every item on it exists because skipping it once cost someone a result.

Two weeks before

Confirm the sending account is active, with enough capacity for the full list plus a retry.
Provision sender numbers in the district's area codes. Plan on enough numbers to keep each one near 50 drops a day for the expected volume.
Verify each callback destination phone with the platform, and test that it rings.
Build one routing profile per candidate or program. Never let two campaigns share sender numbers.
Decide what answers the callbacks: the candidate, a staffed line, or an AI agent with a transfer option.
Have counsel review the script and the consent position for the list, especially for any artificial or cloned voice.

Script and recording

Open with the candidate's name and the office, in the first five seconds.
Identify the campaign committee as required for prerecorded calls, and give a callback number.
One clear action, with the specifics (date, hours, place).
Give the voter a reason to call back.
Under 35 seconds.
Transcribe the final audio and compare it word for word with the approved script before upload.

The list

Pull the freshest file possible, ideally the morning of the drop for GOTV.
Remove anyone who has already voted.
Keep cell numbers only. Normalize to ten digits. Remove duplicate phones so no household gets the same message twice.
Put seed records (staff and candidate phones) at the top so the team hears the live message early.
Keep the source columns (precinct, segment, vote history, age) so results can be cut by them afterward.
Check the row count and a sample of rows against the source file before upload.

Launch settings

Restrict the sending days. For an election-eve message, allow that day only.
Set the window in the recipient's time zone, inside federal and state calling hours, and end it 30 minutes before any hard stop.
Set the hourly rate so the list finishes inside the window with at least 20 percent headroom.
Turn on area-code matching for sender numbers.
Confirm the recording, the list, the routing profile and the schedule one last time on the review screen.

During the drop

Confirm the seed phones received the message within the first hour.
Check pacing and delivery rate at intervals. A delivery rate far below 55 percent early on signals a list or carrier problem worth pausing for.
Watch the callback log and make sure someone is answering.

After the first pass

Build the retry list from carrier errors, no pickups, not-detected and busy outcomes. Exclude every voter already delivered.
Exclude permanent failures: no mailbox, landline, disconnected.
Run the retry the same day if the message is time-bound, or the next day at a slower pace if it is not.
Hand every voter still unreached to the next channel, sorted by reason.

Reporting

Delivered, by attempt, as a share of voters dialed.
Outcomes by reason, so the campaign knows why voters were missed.
Delivery by segment, age and precinct.
Every callback, matched to the voter record, with repeat callers flagged for a personal return call.
A clear statement of what was billed: delivered messages only.

Sources

Results for the two races come from delivery and callback records for the campaigns described, pulled on the evening of the drop. Delivery rates are voicemails delivered divided by voters dialed. Seed records belonging to campaign staff are excluded from every figure. One callback from a number registered to a candidate, most likely a test of the line, is excluded from all callback figures. Expectation ranges in Part Ten reflect our experience across political and commercial programs and are labeled as such. Industry and regulatory facts are drawn from the sources below. Figures from vendors and call-blocking companies are estimates and are identified as such in the text.

Federal Communications Commission, Declaratory Ruling, Petition of All About the Message, LLC for a Declaratory Ruling, FCC 22-85, CG Docket No. 02-278, released November 21, 2022; published at 87 Fed. Reg. 76426 (December 14, 2022). govinfo.gov/content/pkg/FR-2022-12-14/html/2022-26673.htm
Federal Communications Commission, Declaratory Ruling on AI-generated voices under the TCPA, FCC 24-17, adopted and released February 8, 2024. docs.fcc.gov/public/attachments/FCC-24-17A1.pdf
Federal Communications Commission, "Rules for Political Campaign Calls and Texts," consumer guide. fcc.gov/rules-political-campaign-calls-and-texts
Facebook, Inc. v. Duguid, 592 U.S. 395 (2021), decided April 1, 2021.
Campaign Verify, political vetting for 10DLC registration with The Campaign Registry. campaignverify.org; see also Syniverse, "10DLC for Political Campaigns." syniverse.com/insights/10-dlc-for-political-campaigns
Pew Research Center, "Most Americans don't answer cellphone calls from unknown numbers," December 14, 2020 (survey of U.S. adults, July 13 to 19, 2020). pewresearch.org
Pew Research Center, "The Challenges of Contact Tracing as U.S. Battles COVID-19," October 30, 2020. pewresearch.org
Axios, "Why political campaigns won't stop texting you," February 22, 2024 (reporting 15 billion political texts in 2022). axios.com
The Daily Beast, "Six Billion Texts and Counting," 2024, reporting RoboKiller estimates of 2024 political text volume. Vendor estimate. thedailybeast.com
Allison Dale and Aaron Strauss, "Don't Forget to Vote: Text Message Reminders as a Mobilization Tool," American Journal of Political Science 53, no. 4 (2009).
Donald P. Green and Alan S. Gerber, Get Out the Vote: How to Increase Voter Turnout (Brookings Institution Press, multiple editions), summarizing field experiments on canvassing, phone banks, robocalls, mail and text.
Per Egil Hasvold and Richard Wootton, "Use of telephone and SMS reminders to improve attendance at hospital appointments: a systematic review," Journal of Telemedicine and Telecare 17, no. 7 (2011). pmc.ncbi.nlm.nih.gov/articles/PMC3188816
Ballotpedia, "Deepfake policy in the United States, 2019 to present," count of states with enacted laws on deepfakes in political communications as of January 9, 2026. ballotpedia.org

About Workforce Wave

Workforce Wave builds and runs voice AI for organizations that need to talk to a lot of people well: AI voice agents that answer calls for businesses, healthcare practices, municipalities and nonprofits, and outbound programs like the ringless voicemail work described here. We deliver political programs through a small number of campaign partners who handle strategy and voter data, while we run the delivery, the callback routing and the reporting underneath.

If you are planning a voicemail program, a callback line, or both, we would be glad to talk it through. workforcewave.com