Key Takeaways for AI Calling and Texting Teams
- The "revoke all" rule is gone before it ever took effect. The FCC replaced it in a Report and Order (FCC 26-67) adopted September 30, 2026, and released October 1, 2026.
- The new rules take effect November 25, 2026. That date supersedes the waiver that had pushed "revoke all" to January 31, 2027.
- Informational opt-outs can now be category-specific. A "STOP" to a payment reminder can be treated as an opt-out from payment reminders only, not fraud alerts or appointment reminders.
- Marketing opt-outs are still all-or-nothing. An opt-out from any telemarketing or advertising message revokes consent to all future marketing robocalls and robotexts from that caller.
- You can now designate one or more exclusive opt-out methods, such as a reply keyword or key-press, if you disclose it clearly and conspicuously on every call or text.
- More changes are coming. The FCC is taking comment on a 7-business-day opt-out deadline, mandatory two-way texting, a required "revoke all" option, and affiliate rules. Comments are due November 25, 2026.
For teams running AI voice, SMS, and RCS agents, the hard part is no longer whether to honor an opt-out. It is classifying every message correctly and applying each opt-out at exactly the right scope, in real time, across every channel.
Why the New Opt-Out Rules Are Harder for AI Agents
Category-specific opt-outs sound like relief. For AI agents, they add work. Every outbound call and text now has to be classified, every opt-out has to be scoped, and every channel has to agree, in real time.
An opt-out lands in the wrong lane, and you either keep messaging someone who said stop or cut off alerts they still want.
A single line of marketing turns an informational reminder into a message that contains advertising, and the opt-out widens to all marketing.
A voice agent that never heard about a text opt-out keeps calling about the same category.
Why AI agents handle opt-outs better than keyword filters or people
Keyword filters miss anything that isn't on the list. People understand context but aren't consistent across thousands of conversations. AI agents do both: they understand an opt-out in any words, on any channel, and scope it to the right category every time.
| Keyword filter | Human agent | AI agent | |
|---|---|---|---|
| Understands any phrasing | ✕Exact keywords only | ✓Yes | ✓Yes, including angry or profane replies |
| Reads context: "stop calling about my bill" | ✕No | –Usually | ✓Yes |
| Stops one category, keeps the rest | ✕All or nothing | –Depends on training | ✓Yes |
| Inbound and outbound calls, texts, RCS | –Keyword lists and keypresses | –Calls they handle | ✓Every channel and direction |
| Same result every time, at any volume | ✓Yes | ✕Varies by agent and shift | ✓Yes |
Solving that takes a platform that remembers every contact's consent by category and checks it before every flow, call, and text.
How Plura Builds Opt-Out Compliance Into AI Agents
Category-specific opt-outs are only an advantage if your platform can classify, remember, and enforce them at machine speed. Plura was built for that layer.
Every brand gets its own workspace. Every contact keeps its full history.
In Plura, each workspace is its own database. Two brands run in two workspaces, so data, consent, and opt-outs are controlled at the workspace level. Inside a workspace, every contact record keeps a complete history of everything that has ever happened to it, and AI agents can use context from any previous conversation.
| Payment reminders opt-out | true SMS STOP · 9:03 AM |
| Fraud alerts opt-out | false |
| Appointment reminders opt-out | false |
| Marketing opt-out | false |
- 9:02 AMSMSPayment reminder sent
- 9:03 AMSMSCustomer replied "STOP"
- 9:03 AMAutoTag added and opt-out field set: payment reminders
- 9:04 AMSMSConfirmation sent, no marketing
- 2:15 PMVoiceFraud alert call; AI agent loads prior conversation context
- 6:00 PMJourneyPayment reminder journey: contact skipped
Tags and advanced filters decide who enters each journey
Tag an opt-out once and build advanced filters on top of it. In this example, Jordan opted out of payment reminders, so the payment reminder journey skips Jordan while fraud alerts keep flowing.
opt-out:payment-remindersAND tag is not
opt-out:allconsent:fraud-alertsAND tag is not
opt-out:fraud-alertsAND tag is not
opt-out:allCustom fields are checked before any flow starts
Every contact record can hold any custom field you define, such as an opt-out flag for each category. A Condition node checks those fields before a flow starts, so a contact who opted out never enters it.
Also built in
- Every channel, every direction. Inbound and outbound calls, texts, and RCS run through the same opt-out understanding and the same contact record.
- Multi-model intent detection. Plura runs multiple AI models in parallel during each conversation for variable extraction and decision-making, including recognizing opt-outs that don't use a keyword, even angry or profane ones. You control the detection threshold.
- Our own carrier. Plura Connect, Plura's FCC-licensed VoIP carrier, gives us direct control over number provisioning and routing, with no third-party carrier between your agent and the network.
- Litigator and DNC scrubbing. Plura's Blacklist Alliance integration lets clients add their own API key to scrub contacts before outreach.
- Guardrailed scripts. Informational workflows are built so the agent stays on purpose, which keeps a reminder from becoming a marketing message.
Want to see how your current workflows map to the new rules?
Talk to a Plura specialistHow to Program AI Voice, SMS, and RCS Agents for Opt-Outs
The new rules move the compliance burden from "stop everything" to "stop exactly the right thing." For an AI agent, that means four jobs it must do on every interaction: know what category the message belongs to, recognize an opt-out in any form, apply it at the correct scope across every channel, and never let a stray line of marketing change the category.
We recommend resolving ambiguity to the broadest scope. By rule, any marketing content pulls the opt-out into the all-marketing lane. Only a clear, informational, single-category opt-out stays narrow.
Start with consent. In a February 2024 Declaratory Ruling, the FCC confirmed that AI-generated voices are "artificial" voices under the TCPA (FCC 24-17). AI voice calls are held to the same consent, identification, and opt-out rules as any other artificial or prerecorded voice call. The revocation rules sit on top of that foundation.
AI voice agents
- ✓Catch "take me off your list"✕Wait for a keypress
- ✓Default broad when unsure✕Guess the narrowest scope
- ✓Lock informational scripts✕Let a reminder turn into an upsell
Why we recommend this
- Detect opt-outs in natural language. Unless you've designated an exclusive method, "take me off your list" is a valid revocation. The agent's intent model must catch it as reliably as a key-press.
- Map the scope, and default broad when unsure. "Stop calling me about my bill" points to one category. "Stop calling me" does not. When intent is ambiguous, suppress more, not less.
- Disclose the designated method on every call if you use one, and keep honoring clear verbal opt-outs anyway. It costs nothing and removes a litigation argument.
- Lock informational scripts. Payment, fraud, and appointment workflows should be technically unable to pitch, upsell, or promote.
Cross-Channel Opt-Out Rules for AI Agents
One STOP. One consent record. Every channel updated before the next message.
Why we recommend this
- One consent record per contact, per category. The amended rule describes the category as "informational robocalls and text messages." It doesn't address channels, so the conservative approach is to let a payment-reminder opt-out received by text also stop payment-reminder calls.
- A separate marketing lane. Any opt-out from a marketing message revokes consent to all marketing calls and texts from that caller; as a best practice, apply it across every campaign and vendor.
- Suppress in real time. The ceiling is 10 business days, and the FCC is weighing seven. An AI platform should stop the very next message, not the next batch.
- Keep the evidence. The FCC imposed no new recordkeeping requirement. In a dispute, though, you will want the transcript, timestamp, assigned category, and suppression action for every opt-out.
AI Agent Compliance Checklist: Copy-for-LLM Audit Prompt
Copy this prompt into Claude, ChatGPT, or Gemini, then paste in your scripts, workflows, website copy, or campaign plan. It checks for rule updates first, audits your current setup, and hands back recommended changes, each cited to its official .gov source. Run it before November 25, 2026, and again whenever the FCC acts on the open proposals.
# ROLE You are a TCPA compliance reviewer for AI voice, SMS, and RCS agents. You review the material I paste after this prompt (scripts, workflow logic, consent language, website copy, or campaign plans) and recommend specific changes, each backed by an official U.S. government (.gov) source. # CONTEXT The baseline rules below are current as of November 3, 2026. They come from the FCC's Report and Order FCC 26-67 (effective November 25, 2026), which amended 47 CFR 64.1200. Rules may have changed since then, so you must check before reviewing. # SOURCE POLICY (strict) - Use ONLY official government websites ending in .gov, such as fcc.gov, docs.fcc.gov, ecfr.gov, federalregister.gov, govinfo.gov, and uscourts.gov. - Do NOT use or cite law-firm sites, blogs, vendor or marketing content, news articles, forums, Wikipedia, or any non-.gov page, even if it seems accurate. - Every rule you rely on must be cited to a specific .gov document: document number or CFR section, the paragraph or subsection, the date, and the URL. - Never invent a citation, paragraph number, or URL. If you cannot confirm a citation on a .gov site, write "Citation not verified" instead. - If you cannot verify a rule from a .gov source, label it "Unverified" and do not base a recommendation on it. # STEP 1: CHECK FOR UPDATES (every time this prompt is run) State today's date. If you do not know it, ask me before continuing. If you can browse the web, search .gov sources only for anything published after November 3, 2026 on: - 47 CFR 64.1200, current text: https://www.ecfr.gov/current/title-47/chapter-I/subchapter-B/part-64/subpart-L/section-64.1200 - FCC orders or public notices in CG Docket No. 02-278 (consent revocation) and CG Docket No. 23-362 (AI-generated calls) - Final action on the Further Notice of Proposed Rulemaking in FCC 26-67 (comments due November 25, 2026; reply comments due December 28, 2026) - Any final rule from FCC 24-84 on AI-generated calls and texts - Any federal court decision staying or vacating a rule below For each change found, give the document number, date, what changed, and the .gov URL, then apply it in place of the baseline rules. If you cannot browse, say so plainly and warn that the baseline may be out of date. # BASELINE RULES (as of November 3, 2026) 1. Informational vs. marketing. A call or text is informational only if it contains no advertising or telemarketing. An opt-out from an informational message may be limited to that specific category. An opt-out from any message that contains advertising or telemarketing revokes consent to all future marketing calls and texts from that caller. Source: 47 CFR 64.1200(a)(10); FCC 26-67. 2. Categories. The FCC did not define "category." Its examples include payment reminders, fraud alerts, multi-factor authentication, appointment reminders, and utility outage calls. Source: FCC 26-67. 3. Scope. The rule describes the category as "informational robocalls and text messages." It does not address channels; the conservative reading is that a category opt-out received on one channel covers that category on calls and texts. Treat this as an interpretation, not explicit rule text. Source: 47 CFR 64.1200(a)(10). 4. Opt-out methods. A caller may designate one or more of these as its exclusive opt-out method: an automated, interactive voice or key-press opt-out on a call; a reply text using any of seven words (stop, quit, end, revoke, opt out, cancel, unsubscribe); or a website or phone number the caller designates. The designated method must be clearly and conspicuously disclosed on the call or in the text. Texts may disclose one keyword but must honor all seven. Source: 47 CFR 64.1200(a)(10). 5. No designated method. Opt-outs made by any reasonable means must be honored. Source: 47 CFR 64.1200(a)(11). 6. Timing. Honor every valid opt-out within a reasonable time, not to exceed 10 business days from receipt. Source: 47 CFR 64.1200(a)(10) and (a)(11). 7. Confirmation texts. Optional. A confirmation must only confirm the opt-out and contain no marketing. One sent within 5 minutes is presumed to fall within prior consent. Source: 47 CFR 64.1200(a)(12). 8. One-way texting. Allowed only if each text discloses that replies aren't supported and gives reasonable alternative ways to opt out. Source: 47 CFR 64.1200(a)(10). 9. AI voices. AI-generated voices are "artificial" voices under the TCPA, so AI voice calls need the same consent as other artificial or prerecorded voice calls. Source: FCC 24-17. Baseline source URLs: - FCC 26-67: https://docs.fcc.gov/public/attachments/FCC-26-67A1.pdf - Final rule, 91 FR 67214: https://www.govinfo.gov/content/pkg/FR-2026-10-26/pdf/2026-21821.pdf - Proposed rule, 91 FR 67239: https://www.govinfo.gov/content/pkg/FR-2026-10-26/pdf/2026-21820.pdf - FCC 24-17: https://docs.fcc.gov/public/attachments/FCC-24-17A1.pdf - FCC 24-84: https://docs.fcc.gov/public/attachments/FCC-24-84A1.pdf Pending as of November 3, 2026 (treat as future risk only, unless Step 1 shows they were adopted): a 7-business-day opt-out deadline, mandatory two-way texting, a required one-step "revoke all" option, rules for affiliates and separate lines of business, and proposed AI-call disclosure rules (FCC 24-84). # CHECKLIST [Rule] items are legal requirements. [Best practice] items are recommendations, not legal requirements. Never describe a best-practice gap as a violation. 1. [Best practice] Every AI voice, SMS, and RCS workflow is labeled informational or marketing. 2. [Best practice] Informational categories are defined by consumer purpose, documented, and consistently applied. 3. [Best practice] Informational workflows cannot include offers, promotions, or upsells. 4. [Best practice] Consent is stored per contact, per category, and shared across voice, SMS, and RCS. 5. [Rule] Marketing opt-outs stop all marketing calls and texts from that caller. [Best practice] Suppression extends to every campaign and to vendors calling on the caller's behalf. 6. [Rule] If reply text is a designated method, all seven standardized words are honored; with no designated method, opt-outs by any reasonable means are honored. [Best practice] All seven words and natural-language opt-outs are recognized on every text program. 7. [Best practice] Ambiguous opt-outs default to the broader scope. 8. [Rule] Any designated exclusive method is clearly and conspicuously disclosed on every call and in every text. 9. [Rule] Confirmation texts only confirm and contain no marketing. [Best practice] They are sent within five minutes. 10. [Rule] Opt-outs are honored within a reasonable time, not to exceed 10 business days. [Best practice] Suppression happens in real time. 11. [Best practice] Every opt-out is logged with transcript, timestamp, category, and suppression action. 12. [Best practice] Lists are scrubbed against litigator and DNC databases before outreach. 13. [Best practice] Counsel has reviewed category definitions and exclusive-method disclosures. If my material doesn't give you enough information for an item, mark it "Not enough info" and tell me exactly what to provide. Do not guess. # OUTPUT FORMAT (follow exactly) Write the entire response as one Markdown document: a # title "TCPA Opt-Out Audit, [today's date]", ## headings for each numbered part below, a Markdown table for the findings, and full URLs as Markdown links. If you can create files, also save the response as a downloadable file named tcpa-opt-out-audit-YYYY-MM-DD.md. If you cannot create files, output the Markdown so I can copy it into a .md file. 1. Start with this disclaimer, word for word: "AI-generated review. This may contain errors and is not legal advice. Check every finding against the official .gov sources listed below before acting, and confirm with qualified counsel." 2. Update check: today's date, whether you could browse, and any rule changes found since November 3, 2026, each with its .gov link. If none, say "No changes found" and list the .gov pages you checked. 3. Findings table with these columns: # | Item | Type (Rule or Best practice) | Status (Pass / Gap / Not enough info) | Evidence from my material | Recommended change | Why | .gov source (document, section or paragraph, URL). For Best practice items, write "Recommendation, not a legal requirement" in the Why column and cite a related .gov source only if one directly supports it. 4. Top 3 risks, ranked by severity, each with its .gov citation. 5. Any message that mixes informational content with advertising, quoted from my material, with the suggested rewrite. 6. Any flow where an opt-out could be scoped too narrowly, with the fix. 7. Questions to take to my lawyer. 8. Sources: every .gov page you used, with document number, date, and full URL. # SCOPE This review covers federal FCC rules only. State laws may add requirements. Flag that, but do not state any specific state rule unless you can cite the state's own .gov source. Source of this prompt: Plura, "TCPA Opt-Out Rules for AI Agents: What's New," published November 3, 2026. [add live URL]
- 1Copy the prompt
- 2Paste it into your AI assistant
- 3Add your scripts, flows, or site copy
- 4Get a Markdown report with .gov sources
Prefer the plain checklist?
- Inventory every AI voice, SMS, and RCS workflow and label it informational or marketing.
- Define informational categories by consumer purpose (payment reminders, fraud alerts, appointment reminders), documented and consistently applied.
- Block offers, promotions, and upsells inside informational workflows at the prompt and workflow level.
- Store consent per contact, per category, shared across voice, SMS, and RCS.
- Maintain one marketing suppression list across campaigns and vendors (best practice); extend it to affiliates where consumers would reasonably expect it (affiliate rules are pending in the FNPRM).
- Train opt-out detection on all seven standardized words plus natural-language variations.
- Default ambiguous opt-outs to the broader scope.
- If you designate an exclusive method, disclose it clearly and conspicuously on every call and in every text.
- Keep confirmation texts confirmation-only (rule) and send them within five minutes (best practice; five minutes earns a presumption of consent).
- Suppress in real time (best practice); the rule is a reasonable time, not to exceed 10 business days.
- Log the transcript, timestamp, category, and suppression action for every opt-out.
- Scrub lists against litigator and DNC databases before outreach.
- Have counsel review your category definitions and exclusive-method disclosures.
How AI Agents Must Classify Messages: Informational vs. Marketing
Category-specific opt-outs apply only to informational messages, which the amended rule defines as a call or text that "does not contain any advertising or telemarketing." (FCC 26-67) Everything else stays under an all-marketing opt-out.
Telemarketing opt-outs stay broad. A revocation made in response to a call or text that contains an advertisement or constitutes telemarketing revokes consent to all future advertising and telemarketing calls and texts from that caller. The FCC noted there are no exemptions from the consent requirement for that category. It also reminded telemarketers that, absent a specific request to the contrary, a residential subscriber's do-not-call request extends to affiliated entities the consumer would reasonably expect to be included (§ 64.1200(d)(5)).
"Category" is not defined. The FCC illustrated the problem with payment reminders, fraud alerts, multi-factor authentication, appointment reminders, and utility outage calls, but those are examples, not an approved list. Build categories around what the message does for the consumer, not around internal campaign names, vendors, or channels.
| What the agent sends | Classification | Scope of a "STOP" reply |
|---|---|---|
| "Your payment is due Friday." | Informational: payment reminders | Payment reminders only (your choice) |
| "We noticed unusual activity on your card." | Informational: fraud alerts | Fraud alerts only (your choice) |
| "Your appointment is tomorrow at 10 a.m. Reply C to confirm." | Informational: appointment reminders | Appointment reminders only (your choice) |
| "Your payment is due Friday. Ask about our premium plan!" | Contains advertising | All marketing calls and texts from you |
Plura's recommendation: if a single message carries both purposes, treat an opt-out as revoking all marketing and that informational category. Confirm your own policy with counsel.
Can AI Agents Require One Opt-Out Method?
Yes. From November 25, 2026, you can designate one or more of three opt-out methods as exclusive, as long as you disclose the designated method on the call or in the text.
Conversational AI opt-out: callers say it in their own words, at any point in the call
Any of the seven standardized words, or any reply in the customer's own words
A website or phone number you designate for opt-outs
Full rule details
- One keyword is enough to disclose, but all seven must work. Text senders can satisfy the disclosure by including at least one standardized word in each text, as long as every one of the seven is recognized as a valid opt-out.
- Designate nothing, and "any reasonable means" still applies. Callers without a designated method must process revocations made by any reasonable means. If a consumer produces evidence of a request made some other way, it is presumed valid unless the caller shows otherwise, judged on the totality of the circumstances.
- The deadline is unchanged. Every valid revocation must be honored within a reasonable time, not to exceed 10 business days from receipt. For callers with a designated method, that means requests made through that method.
- One-way texting is allowed, with conditions. A sender using a protocol that cannot receive replies must say so in each text and give reasonable alternative ways to opt out.
- Confirmation texts are allowed, not required. A confirmation must only confirm the opt-out and contain no marketing. One sent within five minutes is presumed to fall within the consumer's prior consent.
- Other agencies' methods don't qualify. The FCC declined to let callers use an opt-out method set by another regulator as their exclusive method.
- Existing telemarketing opt-out mechanisms still apply. Artificial or prerecorded voice telemarketing calls, and calls made under certain exemptions, must still provide the automated, interactive opt-out mechanism required by § 64.1200(b)(3).
What Changed in TCPA Opt-Out Rules for AI Calls and Texts?
Three things changed: informational opt-outs can now be limited to one category, you can designate one or more exclusive opt-out methods, and financial institutions can send fraud alerts to numbers from reliable sources. FCC 26-67 revised 47 CFR § 64.1200(a)(9)(iii)(A), (a)(10), (a)(11), and (a)(12) (FCC 26-67).
| Rule | Before | Now · Nov 25, 2026 |
|---|---|---|
| Informational opt-out | Stops all consent-based calls and texts ("revoke all," adopted 2024, never took effect) | May be limited to that category (caller's choice) |
| Marketing opt-out | Covered by "revoke all" | Revokes all marketing calls and texts from that caller |
| Opt-out method | Any reasonable means; no exclusive method allowed | Designate one or more of three methods, or keep honoring any reasonable means |
| Financial-institution fraud-alert numbers | Only numbers the customer provided | Customer-provided, or from one of three reliable sources |
Rule textCallers may treat an informational opt-out as "applying only to the specific category of informational robocalls and text messages to which the revocation was directed and not to all communications." FCC 26-67
When Do the New TCPA Opt-Out Rules Take Effect?
The amended rules take effect November 25, 2026, 30 days after the order was published in the Federal Register on October 26, 2026 (91 FR 67214). Per the FCC's published order, the change to § 64.1200(a)(10) supersedes the delayed "revoke all" date of January 31, 2027, and the FCC declined a request to extend the effective date up to 12 months after publication.
- Feb 16, 2024FCC releases the 2024 TCPA Consent Order: revocation by any reasonable means, a 10-business-day deadline, and the broad "revoke all" readingFCC 24-24
- Apr 7, 2025FCC's Consumer and Governmental Affairs Bureau waives only the "revoke all" portion, to Apr 11, 2026DA 25-312
- Apr 11, 2025The 2024 revocation rules in § 64.1200(a)(10)–(11), except "revoke all," take effectDA 24-1068
- Oct 29, 2025FCC seeks comment on modifying "revoke all" and proposes letting callers designate an exclusive opt-out methodFCC 25-76
- Jan 6, 2026Bureau extends the "revoke all" waiver to Jan 31, 2027DA 26-12
- Sep 30, 2026FCC adopts the new Report and Order and Further Notice (released Oct 1)FCC 26-67
- Oct 26, 2026Order published in the Federal Register91 FR 67214
- Nov 25, 2026New rules effective; comments due on the open proposals91 FR 67214, 91 FR 67239
- Dec 28, 2026Reply comments due on the open proposals91 FR 67239
What Was the TCPA "Revoke All" Rule?
Under the 2024 order, an opt-out in response to one type of robocall or robotext applied to all future consent-based calls and texts from that caller. A customer who replied "STOP" to a payment reminder could have cut off their own fraud alerts and one-time passcodes. That provision was waived twice and never took effect. The rest of the 2024 revocation rules, including the 10-business-day deadline, has been in force since April 11, 2025 (the confirmation-text rule took effect earlier).
Pending FCC Rules for AI Calling and Texting: What's Next?
The same FCC item includes a Further Notice of Proposed Rulemaking that could tighten the rules again. Comments are due November 25, 2026, and reply comments December 28, 2026, in CG Docket No. 02-278. None of these proposals is law yet.
| Open proposal | What the FCC is asking | What it would mean for AI agents |
|---|---|---|
| Shorter opt-out deadline | Cut the 10-business-day maximum, for example to 7 business days unless the caller shows a longer period is reasonable; possibly different deadlines for callers that designate an exclusive method and those that don't | Batch or manual suppression gets riskier; real-time suppression becomes the only safe design |
| Mandatory two-way texting | Eliminate the option to use one-way texting with an alternative opt-out method, possibly only for marketing texts | Send-only SMS and RCS programs would need reply handling |
| Required "revoke all" option | Allow category-specific opt-outs only if callers also offer a one-step way to stop all consent-based messages, possibly automated | Agents need a global opt-out path alongside category opt-outs |
| Affiliates and lines of business | Clarify how consent and opt-outs apply across affiliates, divisions, and separate lines of business | Multi-brand clients may need shared suppression across brands |
| Phone number scope | Whether an opt-out applies only to the number that revoked, not every number on the account | Consent records may need to live at the number level |
| Other industry requests | Shrink the standardized word list, drop the rebuttable presumption, allow confirmation calls, let confirmation texts ask which category was meant | Keyword and confirmation logic could change again |
TCPA Opt-Out FAQ for AI Agents
Is the TCPA "revoke all" rule still going into effect?
No. The FCC replaced it in FCC 26-67 before it ever took effect. The replacement rules take effect November 25, 2026, superseding the January 31, 2027 date the "revoke all" provision had been delayed to.
If a customer texts STOP to a payment reminder, can my AI agent still send fraud alerts?
Yes, if the payment reminder contained no advertising, fraud alerts are informational, you have valid consent for them, and you choose to treat the opt-out as limited to payment reminders. The rule is permissive: you may still apply the opt-out to everything.
Do category-specific opt-outs apply to marketing messages?
No. An opt-out in response to any call or text that contains an advertisement or constitutes telemarketing revokes consent to all future marketing calls and texts from that caller.
What counts as a "category" of informational messages?
The FCC did not define it. Its illustrations include payment reminders, fraud alerts, multi-factor authentication, appointment reminders, and utility outage calls. Define categories by their purpose to the consumer and apply them consistently.
Can my AI SMS agent require customers to opt out by replying STOP?
Yes, starting November 25, 2026, if you designate reply-text as your exclusive method and disclose it in every text. You must still honor all seven standardized words: stop, quit, end, revoke, opt out, cancel, and unsubscribe.
How fast must an AI agent honor an opt-out?
Within a reasonable time, not to exceed 10 business days from receipt. The FCC is considering reducing that to 7 business days.
Do these rules apply to AI voice agents?
Yes. The FCC confirmed in February 2024 that AI-generated voices are "artificial" voices under the TCPA, so AI voice calls follow the same consent and revocation rules as other artificial or prerecorded voice calls.
Do AI agents have to disclose that they're AI?
Not under a final FCC rule yet. In August 2024, the FCC proposed requiring AI voice calls to disclose AI use at the start of each call, and requiring consent disclosures that cover AI-generated calls and texts (FCC 24-84). As of November 3, 2026, no final rule from that proceeding had been published in the Federal Register. Some state laws impose their own disclosure duties, so confirm requirements with counsel.
Do these rules apply to RCS messages?
The FCC's revocation rules address text messages without a separate category for RCS. The prudent approach is to apply the same consent, keyword, disclosure, and suppression logic to RCS that you apply to SMS.
Do I have to send a confirmation text after an opt-out?
No. It is optional. If you send one, it must only confirm the opt-out with no marketing, and one sent within five minutes is presumed to fall within the consumer's prior consent.
Disclaimer
This article is for general information only and is not legal advice. TCPA obligations depend on your specific messages, consent records, and use case. Consult qualified counsel before changing your consent or opt-out practices.
References
- Federal Communications Commission, Rules and Regulations Implementing the Telephone Consumer Protection Act of 1991, CG Docket No. 02-278, Report and Order and Further Notice of Proposed Rulemaking, FCC 26-67 (adopted Sept. 30, 2026; released Oct. 1, 2026). docs.fcc.gov/public/attachments/FCC-26-67A1.pdf
- Federal Communications Commission, "FCC Votes to Give Consumers More Choice in the Calls They Receive," News Release (Sept. 30, 2026). docs.fcc.gov/public/attachments/DOC-425498A1.pdf
- Statement of Chairman Brendan Carr, FCC 26-67 (Sept. 30, 2026). docs.fcc.gov/public/attachments/DOC-425498A2.pdf
- Federal Communications Commission, Public Draft, "Enhancing Consumers' Ability to Control Calls Received" (Sept. 9, 2026). fcc.gov/document/enhancing-consumers-ability-control-calls-received
- Federal Communications Commission, Strengthening the Ability of Consumers to Stop Robocalls, Report and Order and Further Notice of Proposed Rulemaking, FCC 24-24 (released Feb. 16, 2024). docs.fcc.gov/public/attachments/FCC-24-24A1.pdf
- Consumer and Governmental Affairs Bureau, Public Notice, Effective Date for TCPA Rules on Revoking Consent Set for April 11, 2025, DA 24-1068. fcc.gov/document/tcpa-rules-revoking-consent-unwanted-robocallsrobotexts
- Consumer and Governmental Affairs Bureau, Order, DA 25-312 (Apr. 7, 2025). docs.fcc.gov/public/attachments/DA-25-312A1.pdf
- Federal Communications Commission, Further Notice of Proposed Rulemaking, FCC 25-76 (adopted Oct. 28, 2025; released Oct. 29, 2025). docs.fcc.gov/public/attachments/FCC-25-76A1.pdf
- Consumer and Governmental Affairs Bureau, Order Extending the Effective Date of the TCPA's Consent Revocation Rule, DA 26-12 (Jan. 6, 2026). fcc.gov/document/cgb-extends-effective-date-tcpas-consent-revocation-rule
- Federal Communications Commission, Implications of Artificial Intelligence Technologies on Protecting Consumers from Unwanted Robocalls and Robotexts, CG Docket No. 23-362, Declaratory Ruling, FCC 24-17 (adopted Feb. 2, 2024; released Feb. 8, 2024). docs.fcc.gov/public/attachments/FCC-24-17A1.pdf
- Federal Communications Commission, Notice of Proposed Rulemaking and Notice of Inquiry, CG Docket No. 23-362, FCC 24-84 (adopted Aug. 7, 2024; released Aug. 8, 2024). docs.fcc.gov/public/attachments/FCC-24-84A1.pdf
- Federal Register, "Enhancing the Ability of Consumers To Control the Robocalls," Final Rule, 91 FR 67214 (Oct. 26, 2026). govinfo.gov
- Federal Register, "Modernizing the Ability of Consumers To Control Robocalls," Proposed Rule, 91 FR 67239 (Oct. 26, 2026). govinfo.gov
- 47 CFR § 64.1200, Delivery restrictions, as amended effective Nov. 25, 2026.