Using AI to Write Client Emails, Posts and Newsletters: FINRA and SEC Rules

By The Professions Cloud Editors Updated October 2026 · 5 sources · 3 min read

Prices and facts checked October 13, 2026. We don't accept payment for rankings.

The short answer: you can use AI to draft client emails, newsletters and social posts, but the output is your communication. FINRA's FAQ is explicit that a firm is responsible for its communications whether a person or AI produced them (FINRA Advertising FAQ D.8). On the SEC side, every advertisement remains subject to the Marketing Rule, and the agency has fined advisers over AI claims they couldn't back up.

Why it matters today: with AI writing tools built into CRMs and notetakers, more client-facing text is machine-drafted than ever. The rules haven't changed; the volume has.

What FINRA says

FINRA added two AI questions to its advertising FAQ on May 10, 2024 (FINRA):

FINRA's broader guidance, Regulatory Notice 24-09, makes the same point: its rules are technology-neutral and apply to generative AI as they do to any other tool (FINRA).

What the SEC says

For registered investment advisers, AI-written marketing is subject to the Marketing Rule (206(4)-1) and the antifraud provisions:

• Claims about AI itself: the SEC fined Delphia ($225,000) and Global Predictions ($175,000) in 2024 over AI claims they couldn't substantiate, under the Marketing Rule (Morrison Foerster).
• Performance numbers AI pulls into a draft: gross and net performance rules still apply. SEC staff FAQs updated March 19, 2025 clarified when extracted performance can be shown gross-only, with conditions such as clear labeling and showing total portfolio gross and net with equal prominence (Thompson Hine).
• Enforcement focus: since February 2025, a dedicated SEC unit covers fraud involving AI (SEC).

The common ways AI drafts break the rules

1. Promissory language: "this strategy will protect your retirement."
2. Invented facts: hallucinated statistics, returns or credentials.
3. Testimonial-style text that implies client endorsements you don't have.
4. Unbalanced claims: benefits without risks.
5. Overstating your own AI use: "our AI picks the best funds."

A review workflow for AI drafts

1. Draft with AI, approve as a human. Route retail communications through your normal principal or CCO review.
2. Fact-check every number against your source documents.
3. Strip superlatives and promises.
4. Keep the final version, and the prompt where practical, in your archiving system.
5. Lock down chatbots: approved content only, with logs retained and periodic review.
6. Train staff on what can and can't be sent without review, such as one-off correspondence versus a newsletter.

Guardrails to put in your prompts

Good prompts reduce what reviewers have to fix. Instructions worth adding to any template your team uses:

• "Use only the facts in the notes below. If a fact is missing, write [CHECK] instead of guessing."
• "Don't predict returns or promise outcomes. Mention relevant risks."
• "Don't include performance figures, client names or testimonials."
• "Write in plain English at a grade 8 reading level, under 200 words."

None of this replaces review, but it makes drafts safer to start from. It also keeps confidential client details out of tools your firm hasn't approved, an issue covered in Regulation S-P and AI vendors.

Choosing tools that draft follow-up emails? See the best AI notetakers for advisors, and check how each handles recordkeeping.

FAQ

Do AI-drafted one-off client emails need pre-approval?

Under FINRA rules, correspondence generally needs supervision and review under your procedures, not necessarily pre-approval. Retail communications usually need principal approval. Follow your firm's procedures.

Can I use AI to write social media posts?

Yes, but posts are communications with the public. They need the same review, retention and content standards as anything you write yourself.

Do I need to tell clients a message was written by AI?

Neither FAQ requires an AI label on communications. What matters is that the content is accurate, balanced and supervised.

Changelog: Oct 13, 2026: first published.