AI video for insurance, finance and mortgage marketing: write the disclosure first

Finance, insurance and mortgage marketing is one of the most closely watched kinds of advertising. A video that would be fine for a gym can break a rule for a lender. This page explains how the main US rules apply when a video is made with AI tools, and where the rules I could read end. It is information, not legal or financial advice. If you operate outside the United States, replace each US body below with your own regulator and ask it what applies.

This page draws on the official pages named below, as of 4 November 2026. Nothing here means a regulator approves the use of AI in advertising. None of them says that. They say the existing rules still apply.

Why finance, insurance and mortgage ads are different

Three reasons.

• Numbers matter. A rate, a payment or a return is a claim a customer acts on. Rules often require a full set of terms to appear next to it.
• Videos are short; disclosures are long. A 15 second reel has little room for the sentence the rules require.
• Several regulators can look at the same ad. A product may fall under a securities regulator, a banking or consumer-credit regulator, a state insurance department, and the Federal Trade Commission (FTC) all at once.

Who writes the rules for your product

If you sell	Main rule-makers (US)	Where to look
Securities, as a broker-dealer	FINRA, the Financial Industry Regulatory Authority, and the SEC	FINRA Rule 2210 (communications with the public) and Regulatory Notice 24-09
Investment advice, as a registered adviser	The SEC, the Securities and Exchange Commission, including its Marketing Rule	SEC press releases and staff guidance
Mortgages and consumer credit	The CFPB, the Consumer Financial Protection Bureau, through Regulation Z (Reg Z), plus state regulators	12 CFR 1026.24, the advertising section of Reg Z
Insurance	State insurance departments. The NAIC, the National Association of Insurance Commissioners, writes model rules that states may adopt	Your state department of insurance and your carrier
Anything sold to consumers	The FTC	FTC advertising, endorsement and AI-claim guidance

State mortgage law and state insurance advertising rules vary, so check the rules for your state. For insurance, an NAIC model bulletin on AI has been adopted by a number of states, but descriptions of it say it concerns insurers' own use of AI systems rather than how an ad is made. Do not read it as an advertising rule. Ask your state department of insurance and your carrier or managing general agent (MGA, the firm that handles underwriting for a carrier) what they require for marketing material, including edited or AI-made video.

What "technology neutral" means for AI video

FINRA Regulatory Notice 24-09, published on 27 June 2024, reminds member firms that its rules are technology neutral. Rule 2210 content standards apply whether a communication is written by a person or generated by a tool, and the firm's supervision duties (Rule 3110) cover the use of generative AI tools. The notice says it creates no new requirements. It does not discuss deepfakes. So its message for a video team is plain: the standards you already follow for a printed brochure apply to the clip, and a supervisor must be able to review it.

The FTC says the same thing from the consumer side. Announcing Operation AI Comply in September 2024, it said there is no AI exemption from the laws on the books. In that sweep it targeted AI-related deception, including fake reviews and income claims made about AI products.

What to take from both: if the old rule would have stopped a printed ad, it stops the AI video. The tool changes how fast you can make mistakes, not what counts as one.

Rate, payment and performance claims: write the disclosure first

This is the method I recommend, and it works for any claim that comes with required wording. Write the disclosure before the script. Then design the video around it.

1. List every claim you want to make: a rate, a monthly payment, a return, a saving, a guarantee.
2. For each claim, find the wording your rules require beside it. Ask compliance. Do not guess.
3. Write that wording in full, as the first thing in the script file.
4. Decide where it will sit on screen and in the audio, and for how long. It must be readable, which usually means on screen long enough to read and not tiny.
5. Only then write the hook and the story. If the disclosure does not fit in the length of the clip, change the claim or the format, not the font size.

Here is why this matters for mortgage-style ads. The CFPB's page for section 1026.24 of Regulation Z says that when an advertisement for consumer credit includes certain trigger terms, it must also state other terms. The trigger terms listed there include the amount of a down payment, the amount of any payment, the number of payments and the amount of any finance charge. Where those appear, the ad must state the required down payment, repayment terms and the annual percentage rate (APR, the yearly cost of credit shown as a percentage). The same section bans some misleading claims for loans secured by a dwelling, including implying government sponsorship and describing a rate as "fixed" in a misleading way. The page also sets out a separate approach for television and radio, so read the section itself, and ask counsel, to see which route fits a video you post online.

In plain terms, a 15 second reel that says "payments from $1,200 a month" may be the exact trigger the rule describes, and then the full terms must appear. Many short videos cannot carry that load. The compliant answer is often to leave the number out of the video and send the viewer to a page that has everything.

An AI avatar should never quote a rate

Here is a firm rule I would set in any shop that does this work: no AI avatar, voice clone or text-to-speech voice ever speaks a rate, a payment, a return or a guarantee. Three reasons.

• A generated voice can misread or change a number, and you may not notice on a quick listen.
• A number spoken by a synthetic person looks like a promise from a real one.
• Numbers should come from one approved source, not from a script someone retyped.

Instead, put rates and payments in a compliance-approved template with a locked field: the number is filled from one approved table, shown as on-screen text with its full disclosure, and never changed by the video editor. The voice talks about the idea (what a rate lock is, why a deductible matters). The locked text carries the numbers.

Testimonials and clients that do not exist

An AI-made "satisfied customer" is a fake testimonial. The FTC says its final rule on fake reviews and testimonials, announced in August 2024, bans fake or false testimonials, including those from people who do not exist, and allows civil penalties against knowing violators. For a financial firm there is a second layer, because securities and advisory rules also regulate testimonials. It is unconfirmed how the SEC's Marketing Rule treats an actor or an AI character, and a vendor article that claimed to say so gave rule citations that are unverified. The advice does not depend on that: do not use invented clients. Use real clients who agreed in writing, with every disclosure your rules require, and have compliance approve them.

For the consumer-protection side of testimonials in general, read fake testimonials and AI reviews: the legal risk. For the same ideas applied to law and health, see where AI video is risky for law firms and clinics.

AI-washing: claims about your own AI

AI-washing means saying you use artificial intelligence when you do not, or saying it does more than it does. The SEC's press release 2024-36, dated 18 March 2024, announced settled charges against two investment advisers, Delphia and Global Predictions, for false and misleading statements about their use of AI. They paid $225,000 and $175,000, a combined $400,000, according to the release. The SEC said firms should not mislead the public by saying they are using an AI model when they are not.

That matters for video because marketing teams love the phrase "AI-powered". Run an AI-claims audit before any film goes live:

Claim in the video	Evidence to hold	Pass or fail
"Our AI picks your best policy"	A document showing what the system does, what data it uses, and who checks the result	Fail if the answer is "a person does it with a spreadsheet"
"AI-powered underwriting"	The model, its role in decisions, and compliance review	Fail if you cannot describe it in one plain sentence
"Smarter returns with AI"	Backtested or real results with the required performance disclosures	Fail if it implies a return you cannot support
"This video features our virtual adviser"	A note that it is a synthetic presenter, and who is responsible for the content	Fail if viewers may think it is a licensed person

Every sentence that says AI should have a document behind it. If it does not, cut the sentence.

Safe uses

These are the uses I would start with, subject to your own compliance approval:

• Explainers of concepts: what a deductible is, how a fixed and a variable rate differ in general, what a fiduciary duty is. Animated diagrams, no numbers.
• Captions and subtitles on a video of the real adviser or loan officer.
• Translation of an approved video, checked by a fluent speaker who understands the terms.
• The real adviser reading an approved script on camera, with AI used only for tidying audio and cutting.
• Neutral visuals. Be careful with generated "dream life" scenes, such as a yacht or a beach house, because they can imply outcomes. One vendor article flagged this, but no rule on it is confirmed. I would still avoid it.

Compliance-vetted carrier or lender templates often beat custom AI video for any ad that carries a rate. A real adviser or loan officer on camera is safer than any avatar. For the general question of AI presenters, see talking presenter video without filming yourself and, for wider AI use in an advisory firm, AI for financial advisors.

The review step and the paper trail

FINRA's notice ties AI use to supervision, so you need a record that shows a person reviewed and approved the video. Specific retention periods are unverified (a vendor article quoted one for advisers and it was unconfirmed), so ask your compliance officer how long to keep them. A simple pre-approval log:

• Script version and asset version, with file names and where it will be posted.
• Tools used and what each did, including any AI voice or avatar.
• Claim list with the evidence and the required disclosure for each.
• AI-claims audit result (the table above).
• Disclosure placement: wording, on-screen time, audio.
• Reviewer: name, role, date, and approved or rejected.
• Change control: any edit after approval means a new version and a new review.

A video sent to one prospect by name is a different thing from a broadcast ad under some adviser rules, according to a vendor article that is unverified. Ask compliance whether your one-to-one videos are treated as advertising.

What to ask an AI video vendor

• Where do my uploads and prompts go, and are they used for training? See does your AI tool train on client work?.
• Can I export a record of what was generated and when?
• Does the tool add labels or hidden marks to outputs, and can I add my own disclosure?
• What does the vendor say about my responsibility for the content? (Most AI video terms put it on the user.)
• Will the vendor sign a data agreement that fits my regulator's needs?

If an agency makes your videos, see should agencies disclose AI use? and write the compliance sign-off into the contract.

When to skip AI video

Skip it if the video would need a rate or payment that you cannot show with its full terms, if the point of the video is an invented customer, if you cannot back up a single "AI-powered" claim, or if your carrier or compliance team has not approved it. A plain video with the real adviser, one idea and a link to a page with the full terms is dull, and it is the one that passes review.

How I'd approach this

Start with one concept video: a real adviser or loan officer, an approved script with no numbers, AI used only for captions and clean-up. Write the disclosure first, fill in the sign-off log, and get compliance to approve the exact file. Run the AI-claims audit on your website and your pitch at the same time, since it is cheaper to find an unsupported claim now. Keep rates in locked templates, and never in a generated voice. If you want a repeatable setup, with a locked-field template, an approval step and a log that fills itself in, Blur Studio builds that workflow, and you can see how it works.