What is Cowork? Use Cases Blog Services AI Readiness Quiz FAQ Book a Call

Blog · Compliance

Using Claude or ChatGPT to write marketing copy — what the FTC expects

The 2026 US marketer's guide to AI-generated content and Federal Trade Commission compliance: endorsement guides, the impersonation rule, Section 5, and the state laws stacking on top.

13 November 2026 · 8 min read

The short answer. The FTC does not require a blanket “written by AI” label on marketing copy in the United States. It does require that every claim be truthful, substantiated, and not misleading — and it treats AI-generated fake testimonials, synthetic endorsers, impersonation deepfakes, and unlabeled synthetic imagery as unfair or deceptive acts under Section 5 of the FTC Act. Use Claude or ChatGPT to draft copy freely. Disclose whenever a reasonable consumer would be materially misled without the disclosure.

FTC Endorsement Guides — 16 CFR Part 255

The FTC updated its Endorsement Guides in 2023 and finalized the Rule on Consumer Reviews and Testimonials in August 2024. Together they put AI-generated social proof squarely in enforcement territory.

AI-generated testimonials and reviews

Fabricated reviews are now subject to civil penalties of over $50,000 per violation. That includes AI-written reviews posted under fake personas, AI-generated star ratings, and any “customer quote” invented by a model without a real customer behind it. The rule is intent-neutral: if a reasonable reader would assume the review came from a genuine customer, and it didn't, it's deceptive.

Synthetic influencers and virtual endorsers

Virtual influencers (fully CGI or AI-generated personas) and AI-generated spokesperson content are permitted, but the Endorsement Guides still apply:

  • Material connections to the brand must be disclosed clearly and conspicuously
  • The endorsement must reflect the honest opinion of the sponsor — a synthetic persona cannot claim experiences it never had
  • Performance, health, or results claims made by a synthetic endorser still require the same substantiation as human-presented claims

Rule on Impersonation of Government and Businesses — 2024

The FTC's impersonation rule took effect in April 2024 and gives the agency authority to pursue direct federal court penalties against marketers who impersonate businesses or government entities. The FTC has since proposed extending the rule to cover impersonation of individuals, aimed squarely at deepfakes.

For marketers using generative AI, the risk zones are:

  • AI-generated lookalike branding, logos, color systems, or packaging designed to piggyback on a competitor
  • Deepfaked voices or video of real executives, celebrities, or public figures in advertising
  • AI-generated email templates or landing pages that mimic another company's transactional communications
  • Fake “as seen on” endorsements from news outlets or industry bodies the brand has no relationship with

Penalties can reach $51,744 per violation (2026 adjusted figure) and the FTC can seek consumer redress directly in federal court — faster and sharper than the old administrative route.

Section 5 of the FTC Act — deception and unfairness

Section 5 is the baseline. It prohibits “unfair or deceptive acts or practices in or affecting commerce” and applies to every ad regardless of who — or what — drafted it. For AI-written marketing claims, the three enduring tests are:

  1. Truthful. Every express or implied claim must be factually accurate. Claude or ChatGPT will cheerfully draft specific percentages, timeframes, and comparative claims; the marketer is responsible for verifying each one before the copy runs.
  2. Substantiated. Performance, health, earnings, and comparison claims require competent and reliable evidence on file at the time the ad runs. “The AI wrote it” is not a defense.
  3. Not misleading in net impression. Even technically true claims can be deceptive if the overall takeaway misleads a reasonable consumer. AI copy that stacks hedged true statements into an exaggerated impression is still deceptive.

FTC Chair Lina Khan's successors have continued the “Operation AI Comply” enforcement sweep, and the agency has made clear that marketers — not AI providers — carry the liability for deceptive AI-generated claims.

AI-generated images and video — labeling expectations

There is no single federal rule forcing a “made with AI” watermark on commercial imagery in 2026, but two authorities overlap:

  • FTC. Section 5 prohibits synthetic imagery that materially misrepresents a product, service, or result. An AI-generated before/after image for a cosmetic brand, or a synthetic product demonstration, must either reflect real outcomes or be clearly disclosed.
  • FCC. The FCC requires on-air disclosure of AI-generated content in political broadcast ads and is consulting on wider rules. Marketers running political or issue-advocacy creative should treat disclosure as the default.

Industry signals point the same way: the Content Authenticity Initiative (C2PA) metadata standard and major platform labeling policies on Meta, TikTok, and YouTube mean synthetic marketing imagery is increasingly flagged automatically. A disclosure is cheaper than a takedown.

State overlays — what layers on top of the FTC

State attorneys general enforce consumer protection acts that mirror Section 5 (New York GBL § 349, Massachusetts Chapter 93A, Washington CPA, Texas DTPA). Any FTC-deceptive AI ad is typically state-deceptive too. Four additional state overlays matter for marketers:

  • California AB 2655 & AB 2839 — election-related deepfake rules that restrict materially deceptive synthetic political content and require labeling in defined windows around elections
  • California B&P Code § 17941 (BOT Act) — requires disclosure when a chatbot interacts with California consumers to incentivize a transaction or influence a vote
  • Colorado AI Act (effective February 2026) — risk-management duties for developers and deployers of high-risk AI systems affecting consequential decisions
  • Utah AI Policy Act — requires clear disclosure when consumers interact with generative AI in regulated-service contexts and in response to direct consumer questions

Practical framework — 5 steps for agencies and in-house teams

  1. Treat AI as a drafter, not a publisher. Every AI-generated claim gets a named human reviewer before it ships. Record who reviewed what in your campaign file.
  2. Substantiate before you schedule. If the copy contains a number, a comparison, or a health/earnings claim, the evidence sits in the folder before the ad is scheduled. No evidence, no claim.
  3. Ban synthetic social proof. No AI-written reviews, no AI-generated testimonials, no composite “customer stories” unless clearly labeled as illustrative and no real person is implied.
  4. Label synthetic media when materiality is at stake. Use a plain-English disclosure (“AI-generated image”, “dramatization”) whenever a reasonable consumer could mistake synthetic imagery for a real product shot, customer, or event.
  5. Document your AI usage policy. Keep a short written policy on which tools are approved, what inputs are allowed, and how outputs are reviewed. State AGs and the FTC increasingly ask for this during inquiries.
What Claudable teaches Our US training programs include a compliance module: the prompt patterns that force Claude to flag unverified claims, a review checklist mapped to Section 5, and templates for the disclosure lines that actually satisfy the FTC Endorsement Guides. We build it into the way your team uses Claude every day.

FAQ

Common questions on AI marketing and FTC rules

Do I have to disclose that marketing copy was written by AI?

The FTC does not require a blanket “written by AI” label on every ad or blog post. Disclosure obligations attach to specific claims — fake testimonials, synthetic endorsers, impersonations, and unsubstantiated performance claims. A human-edited, substantiated ad written with Claude's help typically needs no “AI-written” label, but a synthetic customer review does.

Are AI-generated testimonials illegal under US law?

The FTC's updated Endorsement Guides and the 2024 Rule on Consumer Reviews and Testimonials prohibit fabricated reviews, including AI-generated ones that readers would reasonably assume were written by real customers. Penalties can reach over $50,000 per violation. Synthetic influencers must disclose material connections and cannot convey experiences no real person had.

What does the FTC's 2024 impersonation rule mean for AI-generated brand lookalikes?

The Rule on Impersonation of Government and Businesses (effective April 2024) lets the FTC pursue direct federal court penalties against marketers who use AI to imitate another business's branding, voice, logos, or executives in ways likely to deceive consumers. AI-generated deepfakes of real CEOs or lookalike campaigns targeting a competitor's customers are now higher-risk than ever.

Do AI-generated images and video in ads need to be labeled?

There is no single federal labeling rule for AI imagery in commercial ads, but FTC Section 5 still prohibits deception. If a synthetic image could materially mislead consumers (for example, showing a product doing something it cannot), you must either not run it or disclose clearly. FCC rules add disclosure requirements for AI-generated content in political broadcast ads, and several states — notably California — have added election-specific labeling laws.

Does Section 5 of the FTC Act apply differently to AI-written ads?

No — Section 5 applies the same way regardless of who or what drafted the copy. The marketer is responsible for every claim in the ad. Using Claude or ChatGPT does not shift liability to the AI provider. If the copy makes a performance, health, earnings, or comparison claim, the marketer must have the substantiation on file before running it.

Which state laws should US marketers watch alongside the FTC?

California AB 2655 and AB 2839 govern election-related deepfakes. State UDAP and consumer protection acts in New York, Massachusetts, Washington, and Texas mirror Section 5 and can be enforced by state AGs. California's BOT disclosure law requires disclosing when a chatbot interacts with consumers in a commercial context. Colorado's AI Act (effective 2026) adds risk-management duties for high-risk AI systems.

This article is general information, not legal advice. For specific campaigns, consult your advertising counsel.

Train your team to use Claude the compliant way

Claudable's US programs include prompt patterns, review checklists, and disclosure templates that keep your AI-generated marketing on the right side of the FTC.

See Training Options Book a Discovery Call