How to Substantiate AI Marketing Claims and Avoid FTC Enforcement

The new compliance frontier for AI claims

Companies promoting "AI-powered" features face a substantiation problem that traditional advertising rules weren't designed to handle. While regulators still require a reasonable basis for objective claims before they're made public, the complexity of AI systems makes that basis harder to establish, evaluate, and maintain over time.

The gap between marketing language and technical reality has drawn enforcement attention. The FTC recently settled cases against Workado, LLC over claims its AI content detector was "98% accurate" when independent testing showed 53% accuracy. In separate actions, the agency challenged Cox Media Group and two marketing agencies over "Active Listening" services that allegedly used algorithms to monitor smart device conversations—technology the FTC said didn't actually exist. Those settlements totaled nearly $1 million in penalties.

Why it matters

AI claims span marketing, sales materials, RFP responses, and executive communications—creating multiple points where unsupported statements can expose companies to regulatory risk. Unlike static product features, AI systems change as models update and vendor APIs evolve, meaning a claim that was accurate last quarter may no longer hold up. Companies that can't document their AI capabilities face both enforcement action and customer trust erosion.

Define what the claim actually promises

Before gathering evidence, legal and compliance teams need to ask what a reasonable customer would understand from the claim. "AI-powered customer service" could mean a system handles interactions autonomously, or it might mean AI suggests responses that humans review before sending. Samsung's experience illustrates the risk: the National Advertising Division challenged claims that "smart" connectivity in Bespoke refrigerators was AI-driven, leading Samsung to discontinue those statements.

The interpretation question matters because it determines what evidence is required. A "95% accurate" claim needs documentation of how accuracy was measured, what data was used, and under what conditions. A "more accurate than human reviewers" claim requires identifying the reviewers, their tasks, and whether the comparison was statistically valid.

Vendor claims require independent verification

Substantiation becomes more complex when companies rely on third-party AI technology. Retailers, financial institutions, and software companies increasingly incorporate vendor-supplied AI tools, and vendor marketing language often migrates into customer-facing materials.

Vendor marketing materials should not become a company's substantiation record, according to Andrew Lustigman and Barry Greenbaum of law firm Olshan, writing for Corporate Compliance Insights. Companies should verify vendor claims before incorporating them into their own communications. Contracts can require vendors to provide technical documentation, testing reports, and notice of material changes—but contractual protections don't substitute for understanding the product.

Maintain a current substantiation record

For material AI claims, companies should maintain a record that identifies the precise claim, where it's used, the product involved, supporting evidence, material limitations, the responsible individual, and the last review date. This living document ensures ongoing accuracy as models and capabilities change.

Before approving any AI claim, organizations should review what they're saying, what customers would reasonably understand, and what evidence exists today. Horizon Brands faced an NAD challenge over its "AI-powered Smart Baby Monitor" that demonstrated the importance of disclosing limitations—while the presence of an AI chip supported basic AI claims, the company had to discontinue claims about ensuring infant safety and note restrictions on emotion and motion detection features.

Cross-department coordination prevents inconsistency

AI claims often originate from multiple departments. Marketing approves website language, product teams describe features in sales presentations, salespeople make broader representations, and executives characterize technology in interviews. Each statement may seem reasonable individually, but together they can create an impression that exceeds what the technology actually does.

A coordinated review process involving legal, compliance, engineering, and business teams helps ensure everyone works from the same understanding of the technology and the evidence supporting claims. The goal isn't identical language everywhere, but preventing material differences in what various communications convey about capabilities.

These details were first reported by Corporate Compliance Insights in an analysis by Lustigman and Greenbaum.