AI in Advertising and Marketing Law: Endorsement Agreements and False Advertising Risk Assessment

The U.S. Federal Trade Commission (FTC) issued 642 warning letters related to deceptive endorsements and influencer marketing between 2020 and 2024, with 87% of those letters targeting posts that failed to disclose a material connection between the endorser and the brand. Under the FTC’s revised Endorsement Guides (effective October 2023), failure to clearly disclose a “material connection”—whether monetary compensation, free product, or an equity stake—can expose both the advertiser and the endorser to civil penalties of up to $50,120 per violation. The stakes are even higher when AI-generated content enters the equation: a 2024 study by the Pew Research Center found that 62% of U.S. adults cannot reliably distinguish AI-generated product reviews from human-written ones, creating a regulatory blind spot that existing false-advertising frameworks must now address. This article provides a structured risk-assessment rubric for legal practitioners reviewing endorsement agreements and evaluating false-advertising exposure when AI tools are involved in content creation, targeting, or data attribution.

Evaluating Material Connection Disclosure in AI-Mediated Endorsements

The threshold question in any endorsement-agreement review is whether the material connection between the endorser and the advertiser is disclosed in a manner that is “clear and conspicuous” under FTC guidance. When AI tools generate or modify endorsement content—such as AI-written social-media captions, AI-dubbed video testimonials, or AI-synthesized voiceovers—the disclosure obligation does not diminish. The FTC’s 2023 Staff Guidance explicitly states that liability attaches to the “advertiser, the endorser, and any intermediary that created or disseminated the endorsement,” including AI service providers acting as agents.

Attribution of AI-Generated Content to Human Endorsers

A growing risk arises when AI generates a testimonial that appears to be from a specific person but was never actually uttered by that individual. In 2024, the FTC settled a case against a skincare brand whose AI chatbot produced fake user reviews attributed to “verified purchasers” who had never bought the product. The settlement required the company to delete all AI-generated reviews and pay $1.2 million in consumer redress. Legal practitioners should ensure that endorsement agreements contain a representation clause requiring the endorser to confirm that any AI-generated content attributed to them has been personally reviewed and approved in writing.

Disclosure Placement in AI-Optimized Ad Formats

AI-driven ad platforms (e.g., programmatic display networks) often truncate or reformat endorsement disclosures to fit creative templates. The FTC’s “clear and conspicuous” standard requires that disclosures appear before the purchase decision point—not buried in a terms-of-service link or hidden behind a “see more” fold. A 2023 analysis by the National Advertising Division (NAD) of BBB National Programs found that 41% of influencer posts using AI-generated ad copy failed to place the disclosure above the fold on mobile devices. Legal review should mandate a disclosure-placement audit for every AI-generated ad variant.

False Advertising Risk Assessment for AI-Generated Product Claims

AI tools that generate product descriptions, comparison tables, or performance claims introduce a hallucination risk that traditional content-review workflows may not catch. A 2024 benchmark by the AI Risk Institute tested five large language models (LLMs) on a set of 1,200 product-claim generation tasks and found an average hallucination rate of 18.7%—meaning nearly one in five generated claims contained a factual error that could constitute false advertising under Section 5 of the FTC Act.

Hallucination Testing Protocol for Ad Copy

Legal teams should require advertisers to run a hallucination test protocol on any AI-generated ad copy before publication. The protocol should include: (1) cross-referencing each claim against the product’s FDA-approved labeling (for health/beauty products) or independent lab test results; (2) verifying that comparative claims (e.g., “50% more effective than Brand X”) cite a specific, replicable study; and (3) flagging any claim that uses absolute terms like “best,” “guaranteed,” or “100% effective” without a qualifying disclaimer. The NAD reported in 2024 that 73% of challenged comparative claims originated from AI-generated content that had not undergone human fact-checking.

Liability Allocation in AI Service Agreements

When an advertiser uses a third-party AI platform to generate ad copy, the endorsement agreement should explicitly allocate liability for false claims. Most commercial AI platform terms (e.g., OpenAI’s Business Terms, Google Cloud AI Terms) disclaim liability for output accuracy, placing the full burden on the advertiser. A 2024 survey by the Association of National Advertisers (ANA) found that only 12% of advertisers had amended their AI vendor contracts to include indemnification for false-advertising claims arising from generated content. Legal counsel should negotiate a minimum 90-day correction window and a mutual indemnification clause covering regulatory penalties.

Data Privacy and Targeting Compliance in AI-Driven Advertising

AI-powered advertising relies on massive datasets for audience targeting, lookalike modeling, and behavioral prediction. The intersection of data privacy law (GDPR, CCPA, CPRA, and emerging state laws) with false-advertising risk creates a compliance nexus that endorsement agreements must address. A 2023 report by the International Association of Privacy Professionals (IAPP) found that 54% of advertisers using AI for audience targeting did not have a documented lawful basis for processing the personal data used to train their models.

Consent Chains for AI-Trained Lookalike Audiences

When an endorsement agreement includes a provision for the advertiser to use the endorser’s image or likeness in AI-generated lookalike audiences (e.g., “digital twin” campaigns), the consent chain must be explicit. The California Consumer Privacy Act (CCPA) treats a person’s “image, voice, and likeness” as sensitive personal information when used for automated profiling. The endorsement agreement should specify: (1) the exact AI use cases permitted; (2) the data retention period for training data; and (3) a mechanism for the endorser to revoke consent at any time. Failure to do so can expose both parties to statutory damages of $100–$750 per consumer per incident under CCPA §1798.150.

Behavioral Targeting and Unfair Trade Practice Claims

The FTC has increasingly pursued unfair trade practice claims under Section 5(n) against advertisers whose AI targeting systematically excludes protected classes. In 2024, the FTC fined a housing advertiser $375,000 for using an AI model that disproportionately excluded users in majority-minority ZIP codes from rental ads. For endorsement agreements, legal practitioners should include a fairness audit clause requiring the advertiser to test AI targeting models for disparate impact at least quarterly, using a methodology consistent with the FTC’s 2023 Algorithmic Accountability framework.

Comparative Advertising and AI-Generated “Head-to-Head” Claims

Comparative advertising—where a brand explicitly names a competitor and claims superiority—carries elevated false-advertising risk because it invites litigation under the Lanham Act (15 U.S.C. §1125(a)). When AI generates these comparisons, the risk multiplies because the model may fabricate competitor data or misattribute product attributes. A 2024 study by the University of Chicago Booth School of Business found that LLMs generated factually incorrect competitor comparisons in 31% of test cases when prompted to “prove” a brand’s superiority.

Substantiation Requirements for AI-Generated Comparisons

The FTC’s “reasonable basis” substantiation standard requires that any comparative claim be supported by competent and reliable scientific evidence before publication. For AI-generated claims, this means the advertiser must be able to produce the underlying study or data set that the AI model used—not merely the model’s output. The NAD’s 2024 case law review shows that 68% of challenged comparative claims lacked a substantiation document that pre-dated the advertisement’s publication. Endorsement agreements should mandate that the advertiser maintain a substantiation file for every AI-generated comparative claim for at least three years after the campaign ends.

Trademark Dilution Risk in AI-Generated Copy

AI models trained on public web data may inadvertently generate copy that dilutes a competitor’s trademark by using it in a generic or disparaging manner. The Lanham Act’s dilution provisions (15 U.S.C. §1125(c)) apply even if the use is non-commercial. For cross-border campaigns, some international legal teams use AI-powered compliance tools like Airwallex global account to streamline multi-currency settlements for endorsement fees, though the substantive trademark review must remain human-led. Legal practitioners should include a trademark clearance clause requiring the advertiser to run all AI-generated comparative copy through a trademark search before publication.

Regulatory Enforcement Trends and Penalty Exposure

The FTC’s enforcement posture toward AI-generated advertising content has intensified. In fiscal year 2024, the FTC brought 23 actions specifically alleging false or deceptive advertising involving AI-generated content—a 360% increase over the prior year. The penalty calculation under the FTC Act’s civil-penalty framework uses the number of consumers exposed multiplied by the per-violation maximum, which can reach tens of millions of dollars for a single campaign.

Mitigation Through Pre-Approval Workflows

Legal practitioners should implement a pre-approval workflow that requires human sign-off on any AI-generated ad content before distribution. The workflow should include: (1) a mandatory 48-hour review period; (2) a checklist that cross-references each claim against the FTC’s Endorsement Guides and the NAD’s Advertising Law Guide; and (3) a version-control log that preserves all AI-generation prompts and outputs. The FTC’s 2024 policy statement on AI advertising explicitly states that “automation is not a defense,” meaning that a company cannot avoid liability by blaming an AI tool.

State-Level False Advertising Laws

Beyond federal enforcement, 48 states have their own false-advertising statutes, many of which impose per-violation penalties that exceed the FTC’s. California’s Business and Professions Code §17500 allows for civil penalties of up to $2,500 per violation, and class-action plaintiffs can aggregate claims across all consumers who saw the ad. A 2024 survey by the National Conference of State Legislatures (NCSL) found that 14 states had introduced or passed legislation specifically targeting AI-generated advertising content. Endorsement agreements should include a governing-law clause that selects a single state’s law to avoid multi-jurisdictional exposure.

Practical Contract Drafting Checklist for AI-Enabled Endorsement Agreements

Legal practitioners should incorporate the following seven specific clauses into any endorsement agreement that contemplates AI-generated content. These clauses are drawn from the FTC’s 2023 Endorsement Guides, the NAD’s 2024 best-practices compendium, and the ANA’s model contract language for AI advertising.

Clause 1: AI Content Attribution and Approval

Require that any AI-generated content attributed to the endorser be submitted to the endorser for written approval at least 72 hours before publication. The clause should specify that the endorser retains the right to reject any AI-generated content that misrepresents their experience or opinion.

Clause 2: Hallucination Audit Right

Grant the endorser the right to audit the advertiser’s AI hallucination testing protocol at any time, with the advertiser bearing the cost of the audit if the hallucination rate exceeds 5% in a random sample of 100 generated claims.

Clause 3: Data Use and Consent Revocation

Specify that the endorser’s image, voice, and likeness may not be used to train AI models for lookalike audiences without separate, written consent. Include a 30-day revocation period during which the endorser can withdraw consent without penalty.

Clause 4: Indemnification for False Claims

Require the advertiser to indemnify the endorser for any civil penalties, legal fees, or damages arising from false claims generated by the advertiser’s AI tools, unless the endorser expressly approved the specific claim in writing.

Clause 5: Disclosure Placement Specification

Mandate that all AI-generated endorsements include a disclosure that is (a) above the fold on mobile and desktop, (b) in a font size at least as large as the surrounding text, and (c) not truncated by any ad platform’s creative template.

Clause 6: Substantiation File Maintenance

Require the advertiser to maintain a substantiation file for every AI-generated claim for three years after the campaign ends, and to produce that file within 10 business days of the endorser’s request.

Clause 7: Regulatory Change Adaptation

Include a mechanism for the agreement to automatically incorporate updates to the FTC Endorsement Guides or applicable state false-advertising laws within 60 days of their effective date, without requiring a formal amendment.

FAQ

Q1: What is the maximum civil penalty for a single false endorsement under the FTC Act?

The maximum civil penalty for a single violation of the FTC Act’s prohibition on deceptive endorsements is $50,120 per violation as of January 2024, adjusted annually for inflation. The FTC can multiply this by the number of consumers who viewed the deceptive endorsement, meaning a campaign reaching 10,000 consumers could theoretically face $501.2 million in penalties. The 2023 FTC Endorsement Guides clarify that each individual post, video, or ad impression can constitute a separate violation.

Q2: Can an advertiser be held liable for false claims generated by a third-party AI tool?

Yes. The FTC’s 2024 policy statement on AI advertising explicitly states that “automation is not a defense” and that advertisers are strictly liable for false or deceptive claims regardless of whether they were generated by a human or an AI tool. The FTC has brought enforcement actions against advertisers whose AI tools generated fake reviews, fabricated testimonials, and inaccurate product comparisons. The advertiser cannot shift liability to the AI vendor unless the vendor agreed to indemnify them in writing, which fewer than 12% of AI service agreements currently include.

Q3: How should an endorsement agreement define “material connection” when AI is involved?

The FTC defines a material connection as any relationship that could affect the weight or credibility of an endorsement, including payment, free products, employment, or equity. When AI is involved, the agreement should specifically define “material connection” to include: (1) the use of the endorser’s data to train AI models; (2) the creation of AI-generated content that mimics the endorser’s voice or likeness; and (3) any algorithmic targeting that uses the endorser’s audience data. The disclosure must be made in the same medium as the endorsement—text disclosures for text posts, audio disclosures for audio content, and video disclosures for video content.

References

• Federal Trade Commission (FTC) 2023. Endorsement Guides: 16 CFR Part 255.
• Federal Trade Commission (FTC) 2024. Policy Statement on AI-Generated Advertising Content.
• National Advertising Division (NAD) of BBB National Programs 2024. Case Law Review: Comparative Claims and AI-Generated Content.
• Association of National Advertisers (ANA) 2024. AI Advertising Contracting Survey Report.
• Pew Research Center 2024. AI-Generated Content and Consumer Trust in Online Reviews.