A financial firm can use AI to help research a topic, outline a page, or produce a rough first draft, but the facts, the judgment, and the voice have to come from the firm, and the result goes through the same review as any other marketing. FINRA has said its communications rules apply whether a person or a tool produced the copy, and the adviser, CPA, and tax rules discussed below judge what a page says, not who drafted it. Treat AI as a drafting assistant, never as a source.
The appeal is obvious. A small firm has more topics than time, and a tool can produce a passable article on retirement income or small business deductions in under a minute. The trouble is that a passable article can contain an invented statistic, a figure from the wrong tax year, or a claim about the firm that nobody approved, all written with the same confidence as the accurate parts.
This is a content-planning question for owners and marketing leads at financial, accounting and tax firms, not legal or compliance advice. Your compliance officer, counsel, or state board is the authority on what your firm may publish.
The rules apply to the words, not the tool
FINRA said this directly in Regulatory Notice 24-09, published June 27, 2024: "FINRA intends for its rules and guidance to be technologically neutral," and "the content standards of Rule 2210 (Communications with the Public) apply whether member firms' communications are generated by a human or technology tool." That notice is addressed to FINRA member firms.
For SEC-registered investment advisers, the marketing rule's general prohibitions apply to the advertisement itself. One of them bars an advertisement from including a material statement of fact for which the adviser lacks "a reasonable basis for believing it will be able to substantiate upon demand by the Commission." A sentence a tool generated is still a sentence the adviser has to be able to support. State-registered advisers should put the same question to their state's rules.
Attorneys, CPAs, and enrolled agents who practice before the IRS face a similar test under Circular 230 (31 CFR Part 10), whose solicitation section prohibits, in any IRS matter, communications containing "a misleading or deceptive statement or claim." CPA firms answer to their state boards and, for AICPA members, to the AICPA Code of Professional Conduct, which "applies to all members." Bring AI-assisted copy to the same reviewer who checks everything else, because the question is whether each statement is accurate and supportable, not who typed it.
Where AI drafts go wrong in financial content
The errors tend to fall into a few patterns, and a reviewer who knows them can move faster.
- Figures without a source or a date. A tool may supply last year's limit or deadline, or a number from nowhere, with nothing to show where it came from.
- Credential slips. A draft may call an enrolled agent "certified," a word Circular 230 section 10.30 says enrolled agents may not use to describe their designation.
- Claims about the firm. Phrases like "our clients typically save" or "we specialize in" appear because they sound natural, not because the firm can support them.
- Invented stories. A tool asked for an example may produce a client anecdote that never happened.
- Rules applied to the wrong firm. A draft may describe the SEC marketing rule as if it covered every adviser, though the SEC's guide says it applies to advisers "registered or required to be registered with the Commission."
- Advice to the reader. A draft meant to explain a service can drift into telling the reader what to do with their own money or return.
Each of these reads smoothly. A fluent paragraph is not a reviewed one, and the smoothness is part of the risk, because a reviewer skimming for tone can miss a number nobody checked. Ask whoever prepares the draft to mark every figure, credential, and claim about the firm with its source before it reaches compliance.
What AI can reasonably do
Google's guidance on generative AI content says the tools "can be particularly useful when researching a topic, and to add structure to original content." The same page cautions against "using generative AI tools or other similar tools to generate many pages without adding value for users," and says "It is critical to manually factcheck and review all AI-generated content for accuracy."
| Task | Reasonable use of AI | What the firm supplies | Who checks |
|---|---|---|---|
| Research | Gathering questions and primary sources to read | Which questions matter to its clients | The writer, against each primary source |
| Outline | Proposing a structure for the page | The actual services, process, and limits | The service owner |
| First draft | Turning approved notes into prose | Approved facts, examples, and wording | Service owner, then compliance |
| Figures and rules | Not a source | A link to the primary source and its date | The reviewer, against that source |
| Client examples | Not a source | A labeled fictional example or an approved real one | Compliance |
| Final copy | Light editing only | Final approval | Compliance or the designated reviewer |
The pattern is simple. AI can arrange what the firm knows, but it cannot be the place the firm's knowledge comes from.
Keep the inputs with the output. Save the notes, source links, and approvals for each AI-assisted page alongside the final version. If someone asks a year later where a figure came from, the answer should be in the file, not in someone's memory of a chat session.
Fictional example: an adviser's retirement income article
Consider a fictional SEC-registered adviser in Missouri. A marketing coordinator asks a tool for an article on how the firm helps clients plan retirement income. The draft is readable. It also says the firm's clients "retire with more confidence than the national average," cites a survey with no link, includes a story about a retired teacher, and ends with a paragraph telling readers how much to withdraw each year.
The coordinator sends it back through a short process. The survey claim and the comparison are cut, since nobody at the firm can substantiate them. The teacher story becomes a labeled fictional scenario that the chief compliance officer approves, and the withdrawal paragraph becomes a description of how the firm's planning meetings work.
That last change needed a person, not a prompt. Since the writer should not have to invent how those meetings run, the coordinator spends twenty minutes with the lead advisor and uses her explanation, in her words. The final draft goes to compliance with each factual statement tagged to its source, and the approved version is kept with the firm's advertising records, since the SEC's guide says advisers "must make and keep copies of all advertisements they directly or indirectly disseminate."
Keep the voice and the review inside the firm
Even an accurate AI draft tends to sound like every other firm's. Prospects choosing between two CPA firms or two advisers often compare how each one explains things, and generic prose gives them nothing to compare. A practical question for the partners is, "If compliance asked where this sentence came from, could we point to a document or a person?" If the answer is no, the sentence is not ready.
That is why Hudson South works from the firm's own material. The Authority Asset gathers approved business knowledge and evidence, such as service descriptions, disclosure documents, and fee schedules. Hudson Twenty-One records how the firm communicates across 21 dimensions so the writing sounds like it. Market Intelligence researches the questions prospects ask and the market around them, and AEO Creation turns those inputs into answer-first content.
For more on keeping drafts sounding like the business behind them, see AI content that still sounds like your business. Advisers working through review should also read what advisers bring to compliance under the SEC marketing rule, and any page that names a credential should follow the guide to credentials a client can check.
Find out where generic content is costing you
On a free 30-minute call, Hudson South walks through where your site and your competitors' sites differ in answering the questions your prospective clients ask.
Get your free competitor analysisSources
- Regulatory Notice 24-09: FINRA Reminds Members of Regulatory Obligations When Using Generative Artificial Intelligence and Large Language Models. FINRA. Read 2026-10-15.
- 17 CFR 275.206(4)-1, Investment adviser marketing. eCFR, National Archives. Read 2026-10-15.
- Investment Adviser Marketing. U.S. Securities and Exchange Commission. Read 2026-10-15.
- 31 CFR 10.30, Solicitation. eCFR, National Archives. Read 2026-10-15.
- 31 CFR Part 10, Practice Before the Internal Revenue Service. eCFR, National Archives. Read 2026-10-15.
- AICPA Code of Professional Conduct. AICPA. Read 2026-10-15.
- Google Search's guidance on using generative AI content on your website. Google Search Central. Read 2026-10-15.