The Autodialer Definition Everyone Gets Wrong

If you ask most business owners, they'll tell you an autodialer is any system that dials numbers automatically. That intuition is wrong, and the gap between intuition and the actual law is where expensive mistakes happen.

The controlling definition comes from the TCPA, enacted in 1991, long before anyone imagined modern calling platforms. Under 47 U.S.C. § 227(a)(1), an automatic telephone dialing system is equipment with the capacity to store or produce telephone numbers using a random or sequential number generator, and to dial such numbers, according to the Congressional Research Service. That generator clause is the part everyone skips.

For years, courts couldn't agree on it. The Second, Sixth, and Ninth Circuits read the definition broadly, while the Third, Seventh, and Eleventh read it narrowly. The Supreme Court settled the split on April 1, 2021, in Facebook, Inc. v. Duguid, unanimously and decisively.

• The Court held that "using a random or sequential number generator" modifies both "store" and "produce," applying the series-qualifier canon to the statute's grammar and punctuation.
• Equipment that simply dials numbers from a stored customer list, without any random or sequential generation, does not qualify as an autodialer under the TCPA.
• The Court rejected the broader reading because it would sweep in "virtually all modern cell phones," which store and dial numbers too.
• Justice Sotomayor wrote that a broad definition would "take a chainsaw to these nuanced problems when Congress meant to use a scalpel."

So a predictive-style system calling through a customer database is probably not an autodialer post-Duguid. Probably. The FCC still interprets the definition more broadly to include predictive dialers, and what "capacity" means remains unsettled across circuit courts. The ground keeps shifting.

Here's why the precision matters. TCPA statutory damages run $500 to $1,500 per call, with no cap. A single non-compliant campaign of 10,000 calls creates potential exposure of $5 million to $15 million. When the stakes look like that, "probably not an autodialer" is not a compliance strategy.

There's a second trap, and it catches more AI callers than the autodialer question does. The FCC's February 2024 Declaratory Ruling confirmed that AI-generated voices are "artificial or prerecorded voices" under the TCPA. That means prior express consent is required for AI voice calls even if your dialing system is not technically an ATDS. The voice itself triggers the obligation. An established business relationship may exempt manual calls from Do-Not-Call rules, but it does not exempt AI voice calls from consent requirements.

This is why My AI Call Center treats every AI-generated voice as an artificial voice requiring prior express consent, and why list source and consent records get reviewed before any campaign launches. The autodialer definition is narrower than most people assume, but the consent rules around AI voices are wider than most people expect. You need to get both right.

What the Supreme Court Actually Held in Facebook v. Duguid

On April 1, 2021, the Supreme Court handed down a unanimous decision that settled years of conflicting lower-court rulings about what counts as an autodialer. In Facebook, Inc. v. Duguid, all nine justices agreed that to qualify as an automatic telephone dialing system (ATDS), a device must have the capacity either to store or to produce telephone numbers using a random or sequential number generator. The ruling reversed the Ninth Circuit's broader reading and resolved a split among six circuits.

The reasoning turned on grammar. Justice Sotomayor, writing for the Court, applied the "series-qualifier canon," holding that the phrase "using a random or sequential number generator" modifies both "store" and "produce," not just "produce." The statutory text of 47 U.S.C. § 227(a)(1) defines an ATDS as equipment with the capacity to store or produce numbers to be called using such a generator, and to dial those numbers.

What the Court refused to do matters just as much. A broader reading, it warned, would cover "virtually all modern cell phones," which can store numbers and dial them. As the Court put it, that approach would "take a chainsaw to these nuanced problems when Congress meant to use a scalpel." The practical result:

• Equipment that dials from a stored customer list, without random or sequential number generation, is not an autodialer post-Duguid.
• Predictive dialers remain a point of tension, since the FCC reads the definition more broadly than the Supreme Court does.
• The "capacity" question is still unresolved across circuits, so the definition's edges remain contested.

That last point shapes how My AI Call Center builds campaigns. Structured calling against approved, permissioned, or reviewed lists keeps a campaign clearly outside the narrow ATDS definition. There's no random or sequential number generation anywhere in the process.

One caution: Duguid isn't the whole compliance story for AI calling. The FCC's February 2024 Declaratory Ruling confirmed that AI-generated voices count as "artificial or prerecorded voices," meaning consent requirements apply even if your dialer isn't technically an ATDS. Statutory damages run $500 to $1,500 per call with no cap, which puts a single 10,000-call campaign at $5 million to $15 million in potential exposure.

Why the Autodialer Label Isn't Your Main Compliance Problem Anymore

If you run AI-powered outbound calls, the question of whether your dialer is technically an autodialer is no longer the biggest thing keeping you up at night. In February 2024, the FCC changed the math entirely.

The FCC's Declaratory Ruling confirmed that AI-generated voices count as "artificial or prerecorded voices" under the TCPA, the same category as traditional robocalls. That includes voice cloning technologies. The ruling applies to any AI technology that initiates an outbound call using an artificial or prerecorded voice to consumers, according to a regulatory update from Mintz.

Here's why that matters more than the autodialer debate. The Supreme Court's 2021 Facebook v. Duguid decision narrowed the autodialer definition to equipment that stores or produces numbers using a random or sequential number generator. Most modern dialing systems, which simply dial from a stored customer list, fall outside that definition. But consent for AI voice calls is mandatory even if the dialing system is not technically an ATDS. As Henson Legal puts it, the AI-generated voice itself triggers the consent requirement. The "artificial voice" provision bypasses the autodialer question entirely.

The stakes are real. TCPA statutory damages run $500 to $1,500 per call with no cap, which means a non-compliant campaign of 10,000 calls could create $5 million to $15 million in potential exposure. And note one trap: an established business relationship may exempt manual calls from Do-Not-Call rules, but it does not exempt AI voice calls from consent requirements.

It's also worth clearing up the headlines. The FCC did not ban AI robocalls. Existing robocall laws now apply to AI-generated calls, so compliant programs remain viable. Non-commercial AI calls to cell phones are permitted with prior express consent, and landline calls without consent remain allowed under specific conditions, including no more than three calls per residential line in any consecutive 30-day period and honoring opt-outs, per Elias Law Group.

For practical purposes, this shifts your compliance focus to the list and the consent record:

• Prior express consent is required before an AI voice call, regardless of dialer technology.
• The AI nature of the call should be disclosed, and recipients can ask whether a call is AI-assisted, request a human, or opt out.
• Opt-out mechanisms need to work fast; an automated opt-out must be available within two seconds of the initial message.
• Consent documentation matters most. Legal guidance stresses consent must be clear, direct, and unequivocal, and carefully documented.

This is how we frame campaigns at My AI Call Center: every campaign runs against approved, permissioned, or reviewed lists, with consent records checked before launch, and AI disclosure on every call. The autodialer definition still matters, but the artificial-voice rule is the gate you actually have to pass through. Requirements vary by location, industry, and consent status, so get legal guidance specific to your situation before launch.

What's Still Unsettled: Capacity, the FCC's Broader Read, and State Mini-TCPAs

Even with the Supreme Court's unanimous 2021 ruling, the autodialer question is not fully settled. Three areas remain genuinely unsettled: what "capacity" actually means, how far the FCC's broader reading can stretch, and how state-level rules layer on top of federal law.

The "capacity" circuit split is the most immediate open question. The Supreme Court decided what counts as an autodialer, but did not define what it means for equipment to have the "capacity" to autodial. According to Congressional Research Service analysis, the Second, Third, and D.C. Circuits treat capacity as a snapshot of the device's capabilities at the time of the offending call. The Seventh, Ninth, and Eleventh Circuits take a functional view, asking what the equipment actually does rather than what it could theoretically do. The same system could be classified differently depending on which circuit hears the case.

The FCC and the Supreme Court also continue to disagree on the definition's outer edge. The FCC has historically read the autodialer provision to capture predictive dialers and other modern equipment, reasoning that technological change has made random or sequential number generation obsolete. The Court rejected that expansion in Facebook v. Duguid, noting that a broader reading would sweep in "virtually all modern cell phones." That tension is unresolved, and the FCC's 2015 attempt to define capacity as including "potential functionalities" was already invalidated by the D.C. Circuit in ACA International v. FCC as creating an "eye-popping sweep."

A third shift arrived in February 2026. In Bradford v. Sovereign Pest Control of TX, Inc., the Fifth Circuit rejected the FCC's prior-express-written-consent rule, holding the TCPA's text does not distinguish between written and oral consent. Per Holland & Knight's analysis, this is among the first TCPA decisions to apply the Supreme Court's 2024 Loper Bright ruling, which ended Chevron deference to agency interpretations. The ruling applies only within the Fifth Circuit and concerns consent form rather than the autodialer definition itself, but it signals broader judicial willingness to reject FCC interpretations that exceed the statute's text.

State "mini-TCPA" statutes add a parallel layer. Several states restrict "automatic dialing and announcing devices" (ADADs), and some proposed frameworks would classify certain voice AI uses as regulated automated decision-making tools. Because these rules vary by jurisdiction and continue to evolve, structured campaigns against approved, permissioned, or reviewed lists, with consent records checked before launch, keep operations clearly inside documented permission and reduce exposure to every layer of this fragmented landscape.

For organizations running AI-powered outbound calls, the practical takeaway is that the autodialer definition matters less than it once did. The FCC's February 2024 Declaratory Ruling made AI-generated voices an independent consent trigger, so compliance turns on documented prior express consent, AI disclosure, and rapid opt-out handling, regardless of whether the dialer technically qualifies as an ATDS. My AI Call Center's practice of reviewing list source and consent records before any campaign launches, declining bought lists without clear permission, and treating AI voices as artificial voices under the TCPA lines up directly with where the law is actually heading.

How to Run AI Calling Campaigns Without Walking Into Autodialer Risk

A single non-compliant campaign of 10,000 calls creates potential statutory exposure of $5 million to $15 million, with TCPA damages of $500 to $1,500 per call and no cap. That math should shape every calling decision you make.

The good news is that structured, consent-based AI calling stays well clear of autodialer risk when you follow a few hard rules. The Supreme Court's unanimous 2021 ruling in Facebook v. Duguid narrowed the autodialer definition to equipment that stores or produces numbers using a random or sequential number generator. Dialing from a real, reviewed list doesn't fit that definition.

Here's the operating discipline that keeps campaigns defensible:

• Structured campaigns against approved, permissioned, or reviewed lists only. Never random or sequential number generation. This keeps you outside the narrow ATDS definition and away from the unresolved "capacity" ambiguity that still splits the circuits.
• Documented consent before launch. The FCC's February 2024 Declaratory Ruling treats AI-generated voices as "artificial or prerecorded voices," so prior express consent applies even when the dialing system isn't technically an autodialer. Consent must be clear, direct, and independently verifiable.
• AI disclosure on every call. Recipients should be able to ask if the call is AI-assisted, request a human, or opt out. Don't bury the disclosure.
• Fast, honored opt-outs. An automated opt-out mechanism should be delivered within two seconds of the initial message and activatable by voice command or key press. Log every opt-out and honor it immediately.
• List source and consent records reviewed before anything dials. Bought lists without clear permission records get flagged or declined. If a list can't support the campaign, say so before anyone spends money.

My AI Call Center runs its managed campaigns on exactly this discipline: one clear goal per campaign, a list and consent review before launch, script and disclosure approval before a single call goes out, and opt-out and DNC logs delivered with every outcome report. The company reports what actually happened and declines lists that won't hold up.

One more thing worth watching: the regulatory floor is still moving. The Fifth Circuit recently rejected the FCC's prior-express-written-consent rule, holding the TCPA's text doesn't distinguish between written and oral consent. That ruling applies only within the Fifth Circuit, and other jurisdictions still follow the FCC framework. Document consent carefully either way, and get legal guidance for your specific situation before you launch.

Ready to run structured AI calling campaigns against approved, permissioned lists? From 9¢ per connected minute, with the full number known before you approve launch. Plan My Campaign.