FTC compliance for AI social media agents

Social agents post for your brand, reply to customers and engage with reviews. The FTC's Endorsement Guides require clear disclosure of material connections. Its 2024 rule bans fake reviews and fake followers. A false claim about a competitor can be defamation. Section 230 doesn't protect what your own agent writes.

Public reply · Brand accountPreview: coming soon

A social agent is replying to a customer who asked how your blender compares with a rival's.

Great question! Ours has a 5-year warranty, and BrightBlend's motors are known to catch fire.

What Cerata will check

  • Passed: Warranty claim matches your published terms
  • Failed: Claim about a named competitor is backed by a source
  • Needs a person: A factual claim about a real company that could be defamatory
Hold

A person on your team decides. A damaging factual claim about a named company needs proof. If it's false, your company answers for it, not the platform.Fix: keep the warranty point and drop the competitor claim, or link the public recall notice that supports it.

Coverage

  • Coming soon:endorsement disclosures, fake reviews and followers, and claims about real companies

HoldAgent's reply

Ours has a 5-year warranty, and BrightBlend's motors are known to catch fire.

Would clearAfter review

Our blender has a 5-year warranty. Here are both spec sheets side by side.

The laws that apply

A brand's posts are advertising. The FTC rules for ads, reviews and endorsements apply, and so does ordinary state law on false statements.

  1. FTC Endorsement Guides16 CFR Part 255, § 255.5

    Coming soon

    Disclose, clearly and conspicuously, any connection between an endorser and the brand that the audience wouldn't expect, such as payment, free products or employment. Its examples cover brand reposts and employee posts.

    16 CFR Part 255 on eCFR (opens in a new tab)
  2. Consumer Reviews and Testimonials Rule16 CFR §§ 465.2, 465.8

    Coming soon

    Bans reviews and testimonials that misrepresent whether the reviewer exists or used the product, and buying or selling fake indicators of social media influence, such as bot followers or views, for a commercial purpose.

    16 CFR Part 465 on eCFR (opens in a new tab)
  3. DefamationState law; New York Times Co. v. Sullivan, 376 U.S. 254 (1964)

    Coming soon

    A company can sue over a false factual statement that harms its reputation. Opinion is generally protected; a false factual claim about a named rival is not. Public figures must also prove actual malice.

    New York Times v. Sullivan on Cornell LII (opens in a new tab)
  4. Michigan AI disclosure in political adsMich. Comp. Laws § 169.259

    Coming soon

    Paid political ads in Michigan with AI-generated images, audio or video must clearly say they were generated in whole or substantially by AI. Several other states have their own rules.

    MCL 169.259 on the Michigan Legislature site (opens in a new tab)

Posts a legal check would sort

A social agent writes fast and in public. These are the kinds of posts a legal check would catch before they go up.

  • Reposting a creator's paid review without saying it's a paid partnershipFlag

    The brand's repost has to disclose the relationship too. Add “paid partner”.

  • Buy 10,000 followers before the launchBlock

    Buying fake followers for a commercial purpose is banned.

  • BrightBlend's blender was recalled last yearHold

    A factual claim about a named rival. A person checks the recall notice first.

  • Thanks for the feedback. Send us your order number and we'll fix it.Clear

    Nothing to disclose, nothing to prove.

Illustrative. Brand names are made up.

What can go wrong

A post can be screenshotted and shared before anyone reviews it, and a social agent posts all day.

$53,088

maximum civil penalty per violation of the FTC's reviews rule, which covers fake reviews and fake followers.

16 CFR § 1.98
$250–$1,000

civil fine per political ad in Michigan without the AI disclosure: up to $250 for a first violation, up to $1,000 after that.

Mich. Comp. Laws § 169.259

Figures from the FTC's inflation-adjusted amounts and the statute. Defamation damages are set case by case under state law. General information, not legal advice.

How Cerata fits

When these checks arrive, they'll use the same step as every other Cerata check: one call before a post or reply goes up.

  1. Before posting, your agent sends the draft and the account it will post from.
  2. Cerata will check disclosures, review rules and claims about real people and companies.
  3. Risky claims wait for a person on your team, and the receipt records who decided.
What your agent gets backPOST /v1/preflight
→ 200 OK
{
  "decision": "hold",
  "fixes": [{ "text": "Claim about a named competitor without a source. Drop it or link the evidence, then post." }],
  "receipt": "eyJhbGciOiJFZERTQSJ9…"
}

Questions

More in the glossary and on the blog.

Does Section 230 protect posts my AI agent writes?

No. Section 230 protects you from liability for content someone else provides. Posts, replies and reviews your agent writes for your brand are your content, so advertising and defamation law apply as usual.

Do brand posts written by AI need a disclosure?

The FTC's Endorsement Guides are about relationships, not tools. If a post endorses a product and the poster has a connection the audience wouldn't expect, such as payment or employment, say so clearly. Some states also require AI disclosures in political ads, such as Michigan.

Can my agent write reviews or buy followers?

No. The FTC's 2024 rule bans reviews that misrepresent whether the reviewer exists or used the product, and buying fake followers or views for a commercial purpose.

Does Cerata check social media agents today?

Not yet. Social media checks are coming soon. Today Cerata checks sales calls and texts (federal rules and Florida), marketing email and rental listings (federal rules). New York City job posts are next.

Start with the checks that are live today.

Social media checks will arrive through the same step.