AI Engine · 5 min read
AI chat, texting and DMs for mortgage: a compliance checklist (TCPA, 10DLC, platforms)
A plain-English checklist for mortgage teams using AI chat, SMS and social DMs: TCPA consent, A2P 10DLC registration, Meta and TikTok rules, disclosure and fair lending.
By Priya Halvorsen (Compliance Content Editor) · Published · Updated
Key takeaways
- Text and call only people who gave prior express consent, and keep a record of it.
- Register business texting under A2P 10DLC; unregistered traffic is filtered by carriers.
- Honor STOP immediately and respect quiet hours in the recipient’s time zone.
- Follow each platform’s messaging rules: Meta response windows, TikTok only where access allows.
- The AI must disclose it is an AI, give the same answers to everyone, and never quote rates or approvals.
AI chat, two-way texting and social DM automation can make a mortgage team dramatically more responsive. They also touch some of the most enforced consumer-protection rules in the US. This checklist explains, in plain English, what a team should have in place before switching AI on across website chat, SMS, Facebook Messenger, Instagram and TikTok. It is general information to discuss with your compliance officer, not legal advice.
What does the TCPA mean for mortgage texting?
The Telephone Consumer Protection Act (TCPA) and the FCC’s rules under it restrict calls and texts made with autodialers or prerecorded or artificial voices without the recipient’s prior express consent (and prior express written consent for telemarketing). Several states have their own, stricter versions. For a mortgage team, the practical rule is simple: no automated texts or calls without consent you can prove.
- Put clear consent language next to every form’s submit button, approved by compliance.
- Save the exact language, the timestamp and the page on the contact record.
- Don’t make consent a condition of getting a quote or a call back.
- When someone texts you first, their message is not blanket consent to marketing; keep replies relevant to their request.
What is A2P 10DLC and why does it matter?
Definition
A2P 10DLC: the US carrier registration system for application-to-person (business) text messages sent from standard 10-digit local numbers. Businesses register their brand and each messaging campaign (use case, sample messages, opt-in method); carriers then deliver registered traffic and heavily filter or block unregistered traffic.
- Register your brand (legal name, EIN, address, website).
- Register each campaign: for example "customer care" and "marketing", with sample messages and how people opt in.
- Show your opt-in language and privacy policy on your website (carriers check).
- Include opt-out instructions in early messages ("Reply STOP to opt out").
- Keep content within your registered use case.
CTIA’s Messaging Principles and Best Practices are the industry reference carriers point to for opt-in, opt-out and content expectations.
What rules apply to Facebook, Instagram and TikTok messages?
| Channel | Key rule | What it means for AI replies |
|---|---|---|
| Facebook Messenger | Meta’s Messenger Platform policies and standard messaging window | Reply to people who messaged you, within the allowed window; no unsolicited promotions outside it |
| Instagram DMs | Meta messaging policies for professional accounts | Same approach; connect only accounts you own |
| Meta ads (housing) | Mortgage ads run in the special ad category for housing with limited targeting | An ads setting your team controls; the AI never changes targeting |
| TikTok messages | Business messaging availability varies by account type and region | Use AI replies only where your account’s messaging access allows; otherwise route to your site |
| Website chat | Your own site and privacy policy | Disclose AI, link privacy policy, consent before texts |
What should the AI itself be allowed to say?
This is where mortgage differs from other industries. An AI that "helpfully" answers the question a borrower actually asks ("what rate can I get?") can create advertising, disclosure and fair-lending problems.
- Disclose: "I’m an AI assistant for…" in the first message, and confirm when asked.
- Answer generally: programs offered, process, documents, hours, from an approved knowledge base.
- Decline specifics: no rates, APRs, payments, approvals, eligibility or advice.
- Be consistent: the same question gets the same answer for everyone, regardless of who they appear to be (fair lending).
- Hand off: to a licensed LO whenever the borrower asks or the topic requires it.
Where conversations end up (illustrative)Illustrative example
- Booked with an LO: 41%
- Handed to a human: 23%
- Answered, no action: 28%
- Opted out: 8%
Show the data table
| Segment | Value |
|---|---|
| Booked with an LO | 41% |
| Handed to a human | 23% |
| Answered, no action | 28% |
| Opted out | 8% |
What about quiet hours and frequency?
The federal TCPA restricts telemarketing calls before 8 a.m. or after 9 p.m. local time, and some states set narrower windows and frequency limits for texts and calls. Build quiet hours into the system by the recipient’s time zone, and cap marketing frequency.
Allowed sending window vs quiet hours under the federal 8 a.m.–9 p.m. rule
Show the data table
| Item | Hours |
|---|---|
| Allowed window | 13 h |
| Quiet hours | 11 h |
The go-live checklist
| # | Item | Owner |
|---|---|---|
| 1 | Consent language on every form, approved and logged | Compliance |
| 2 | A2P 10DLC brand and campaigns registered | Vendor + you |
| 3 | STOP/HELP handling and global opt-out | Vendor |
| 4 | Quiet hours by recipient time zone; frequency caps | Vendor + compliance |
| 5 | AI disclosure in first message and on request | Compliance |
| 6 | Knowledge base approved; no rates, terms or approvals | Compliance |
| 7 | Human handoff rules and response times | Manager |
| 8 | Meta and TikTok accounts owned by the company; policies reviewed | Marketing |
| 9 | Transcript storage and retention defined | You |
| 10 | Weekly transcript review for the first month | Manager |
If you use Mortgage Funnel Suite, these items are part of AI Engine onboarding. If you don’t, the same checklist applies to whichever tools you choose.
Three common scenarios, handled correctly
A borrower texts your number first
Someone texts "do you do FHA?" to a branch number from a yard sign. Replying to their question is responding to their request; the AI can answer generally and offer a call. It should not enroll them in a marketing sequence unless they agree to that, and the first reply should include who you are and how to opt out.
A web form lead with consent
A borrower submits an FHA funnel with the consent checkbox ticked. The consent text, page and timestamp are saved. An immediate acknowledgement text and a follow-up cadence are appropriate, within quiet hours, with STOP honored. If they reply STOP, every automation stops for that number.
An Instagram DM after an ad
A user replies to your Instagram ad. The AI can reply within Meta’s messaging rules, answer general questions and ask whether they would like a call or text. Moving to SMS needs the person’s agreement and the same consent record as a form. The ad itself still needed compliance review and, on Meta, the housing special ad category.
What records should you keep?
If a consent question ever arises, the records are what matter. Keep, for every contact: the consent language they agreed to, when, on which page or channel, any later opt-outs, and the transcripts of automated conversations. Keep templates and knowledge-base versions with their approval dates, so you can show what the AI was allowed to say on a given day. A suite that logs all of this on the contact record and in an audit log saves a great deal of reconstruction later.
Sources
- FCC: Telephone Consumer Protection Act consumer guide
- CTIA Messaging Principles and Best Practices
- Meta Messenger Platform policy overview
- TikTok Business messaging help
- CFPB Regulation Z, 12 CFR 1026.24 (advertising)
Statistics are attributed to the source and date in the text. Examples marked “illustrative” use made-up numbers. This article is general information, not legal, compliance or financial advice.
