The SEC Marketing Rule Doesn't Care If AI Wrote Your Ad
AI can draft your firm's ads and emails. It can't take responsibility for them, and the SEC Marketing Rule already tells you where it will fail.
By Ian Karnell, CEO & Co-Founder, VastAdvisor — November 2026
Yes, an RIA can let AI write its ads and emails, but the SEC Marketing Rule applies to that copy exactly as it applies to copy a person wrote. Rule 206(4)-1 regulates the advertisement, not the author, so the adviser owns every claim the model produces. The practical question isn't whether AI is allowed. It's whether your review process catches the specific mistakes AI tends to make before anything goes out.
Most principals I talk to frame this as a yes-or-no decision: either ban ChatGPT from marketing or quietly let the associate use it. Both are the wrong call. The ban costs you speed your competitors already have. The quiet use creates exactly the record an examiner looks for. The right answer is narrower and more useful: decide which jobs AI does, and build a review step around the failure modes the SEC has already published.
Does the SEC Marketing Rule apply to AI-generated ads?
The SEC Marketing Rule applies to any advertisement your firm disseminates, regardless of whether a person, an agency or a language model wrote it. The rule defines an advertisement by what it does, a communication that offers your advisory services to prospective clients, and lists seven general prohibitions that every ad must clear (Rule 206(4)-1). Nothing in it creates a lower bar for machine-written copy.
FINRA has said the same thing out loud for broker-dealers. Regulatory Notice 24-09 (June 27, 2024) states that FINRA's rules are meant to be technology neutral and that the Rule 2210 content standards apply whether a communication is generated by a human or by a technology tool (FINRA RN 24-09). If you are dually registered, both regimes apply to the same AI draft.
So "the AI wrote it" is not a defense. It's an admission that nobody checked.
What does AI actually get wrong in RIA marketing?
AI gets wrong the same things examiners already flag: confident statements of fact that nobody can substantiate. In April 2024 the SEC's Division of Examinations published its initial observations on Marketing Rule compliance (Risk Alert, April 17, 2024). Read it next to a raw language-model draft and the overlap is uncomfortable. Large language models are trained to produce fluent, plausible, persuasive text. Plausible is not the same as true about your firm.
Here is how common model habits line up with the rule and with deficiencies the staff actually observed:
What a language model tends to do | Rule 206(4)-1 provision at risk | What examiners observed (April 2024 Risk Alert) |
Fills gaps with generic firm descriptions ("our team of specialists") | (a)(1) untrue statement of material fact | Ads saying a network of personnel served clients when one individual did |
Invents or inflates credentials and experience | (a)(1), (a)(2) substantiation | Erroneous personnel qualifications, education and designations |
Describes processes the firm doesn't run ("rigorous screening", "approved list") | (a)(1), (a)(2) | References to screening processes and approved-security lists that did not exist |
Writes reassuring absolutes ("conflict-free", "independent and unbiased") | (a)(1), (a)(3) misleading inference | Ads claiming advisers were "free of all conflicts" when conflicts existed |
Adds social proof and "as seen in" lines | (a)(3), third-party ratings and testimonial rules | "Seen on" national media claims that were paid placements; awards implied as sole top recipient |
Leads with benefits and drops the downside | (a)(4) fair and balanced risks | Social media posts highlighting performance without material risks |
Cites returns or growth figures from training data | (a)(6), (d) performance requirements | Cumulative profit claims and fee assumptions that misstated net results |
The pattern is that a model fills silence with the most typical sentence for a wealth firm. Typical isn't a fact about you. Under (a)(2), every material statement of fact needs a reasonable basis you can substantiate on demand, and the risk alert notes that firms unable to substantiate a claim during an exam were treated as violating that provision too.
The December 2025 follow-up alert focused on testimonials, endorsements and third-party ratings (SEC, Dec. 16, 2025). Those are the other places a model will happily improvise: a sample client quote "for illustration," or a ranking with no methodology disclosure. Neither belongs in a draft that might ship.
Is a personalized AI email still an advertisement?
Usually, yes. The rule's one-on-one exclusion covers communications between an adviser and a single investor, but the SEC's adopting release says communications nominally addressed to one person and actually disseminated to many don't qualify (summary of the adopting release). An AI tool that writes 500 "individual" prospecting emails from one prompt is the modern version of that mail merge. Treat the template and the variations as advertising.
Two more traps sit inside personalization. First, hypothetical performance pulls even a true one-on-one message back into the definition of an advertisement, so a model that helpfully projects "what your portfolio could look like in ten years" has changed the regulatory status of the email. Second, every variant a model produces is a separate piece of copy. If it can rewrite tone, it can rewrite a claim. Your review has to cover the range of outputs, not just the one sample someone approved.
Why are claims about your own AI use now an exam item?
Because the SEC has already fined advisers for overstating it. In March 2024 the SEC settled its first "AI washing" cases against two advisers, Delphia and Global Predictions, for $400,000 in combined penalties (SEC press release 2024-36). Delphia claimed AI and machine-learning capabilities it didn't have. Global Predictions called itself the "first regulated AI financial advisor." Both cases involved the Marketing Rule.
The current exam program keeps this in view. The Division of Examinations' Fiscal Year 2026 priorities say staff will review for accuracy firms' representations about their AI capabilities, and will assess whether firms have adequate policies and procedures to monitor or supervise their use of AI (FY2026 Examination Priorities). The same document lists marketing as a core area in its review of adviser compliance programs.
That creates a quiet irony for firms adopting AI. The moment you start using it, "AI-powered" starts appearing in your own copy. That phrase is a statement of material fact. If your process is a person pasting prompts into a chatbot, you can't call your service AI-driven investment insight and substantiate it.
What should a principal let AI write, and what shouldn't it touch?
Let AI do the work where errors are cheap and visible, and keep it away from anything that states a fact a reviewer can't check from the page. A practical split:
Good jobs for AI (with normal review)
First drafts of educational content on topics you choose: Roth conversion basics, RMD timing, equity compensation explainers.
Rewrites for length and channel: turning an approved article into a LinkedIn post, an email teaser or three ad headline variants.
Tone and reading-level edits on copy that already passed review.
Subject-line and headline testing, where every variant is built from approved claims.
Restricted jobs (AI drafts, compliance sees every version)
Anything describing your services, process, team or fees.
Prospecting emails and ad copy for paid campaigns.
Anything mentioning your own technology or use of AI.
Off limits for AI generation
Performance figures of any kind, including hypothetical and extracted performance. Pull these from your performance records, which Rule 204-2(a)(16) requires you to keep.
Testimonials, endorsements and third-party ratings. These need real people, real disclosures and, for paid arrangements, written agreements.
Statements about regulatory status. The 2024 risk alert flagged ads implying SEC registration signaled skill or approval.
What does a defensible AI marketing workflow look like?
A defensible workflow gives the model only verified facts to work from, screens every output against the rule, puts a named person on the approval, and keeps the record. Six steps:
Build a firm-facts library before you prompt. Approved descriptions of services, fees, team credentials, AUM as of a stated date and conflicts disclosures, all drawn from your Form ADV. The model writes from this, not from its training data.
Write a substantiation note for every material claim. The adopting release suggests a contemporaneous record of the basis for a claim as one way to show reasonable belief under (a)(2). A one-line source per claim is enough.
Screen drafts against the seven general prohibitions. Absolutes, unqualified benefits, implied endorsements and any number all get flagged automatically or by checklist before a human reads for judgment.
Require named human approval. Your CCO or delegate signs off on the final copy and on the variant range for personalized campaigns.
Keep the record. Rule 204-2(a)(11) requires a copy of each advertisement you disseminate, and those records run five years from the end of the fiscal year the material was last disseminated (Rule 204-2). Keeping the prompt, draft, flags and approval alongside the final ad isn't spelled out in the rule, but it is the cleanest way to answer an examiner's "how was this made?"
Put AI in your written policies. The 2024 risk alert cited policies that didn't cover marketing channels firms actually used, such as websites and social media. If AI writes your copy, your Marketing Rule procedures and your annual compliance review should say how.
This isn't legal advice. Confirm your approach with your compliance team or counsel.
How does VastAdvisor approach AI-written marketing?
VastAdvisor was built on the assumption that the model is a drafter, not a decision-maker. Firm Profile Assembly learns your ADV, positioning and compliance context first, and Brand Identity Encoding turns your voice and messaging rules into constraints the copy has to respect. That's step one above, done before a single ad is written.
Every ad and email is then screened before launch against the SEC Marketing Rule, FINRA Rule 2210 and your own policies, with explainable flags, human sign-off and an exportable audit trail. That's the governance layer, not a filter bolted on afterward. Screening supports your compliance review; it doesn't replace your CCO. If you want the longer argument for why governance has to sit inside the system, read our white paper on embedded AI governance in wealth management.
The takeaway
Letting AI write your marketing is a supervision decision, not a technology decision. The rule doesn't change when the author does, the SEC has already published the mistakes it finds, and those mistakes are the ones language models make by default. Give the model verified facts, keep it away from performance and testimonials, put a named person on every approval, and keep the record. Do that and AI makes your marketing faster without making your next exam harder.
Frequently asked questions
Can an RIA use ChatGPT to write marketing content? Yes, but the firm is responsible for every word. The SEC Marketing Rule, Rule 206(4)-1, applies to any advertisement regardless of who or what drafted it. Consumer chat tools don't know your Form ADV, your fees or your conflicts, and they don't keep a review record. Use them for drafts and rewrites from approved facts, then put each piece through your normal compliance review and retain it under Rule 204-2.
Does the SEC Marketing Rule mention artificial intelligence? No. Rule 206(4)-1 is technology neutral and regulates the advertisement, not the tool that produced it. AI shows up elsewhere: the SEC's first "AI washing" cases in March 2024 were brought partly under the Marketing Rule, and the Division of Examinations' FY2026 priorities say staff will review firms' representations about their AI capabilities for accuracy and assess their policies for supervising AI use.
Is an AI-personalized prospecting email an advertisement? Generally, yes. The one-on-one exclusion covers communications between an adviser and a single investor, but communications nominally addressed to one person and actually sent to many don't qualify. An AI campaign producing hundreds of tailored emails from one template is usually advertising. Any email that includes hypothetical performance is an advertisement even when sent to one person. Confirm specifics with your compliance team.
What records should we keep for AI-generated ads? Rule 204-2(a)(11) requires a copy of every advertisement your firm disseminates, kept five years from the end of the fiscal year it was last disseminated, plus records supporting any performance shown under (a)(16). The rule doesn't require keeping prompts, but storing the prompt, draft, compliance flags, substantiation notes and approval with the final ad gives you a complete answer when an examiner asks how it was made.
What is AI washing? AI washing is overstating how a firm uses artificial intelligence in its services or investment process. In March 2024 the SEC settled charges against two advisers, Delphia and Global Predictions, for $400,000 combined over false or misleading AI claims. For RIAs, any statement that your advice or service is "AI-powered" is a material fact that has to be accurate and substantiated under the Marketing Rule.





Comments