AI-generated marketing content disclosed when misleading re: authorship

The FTC's 2024 AI guidance distinguishes between AI as a writing tool and AI-generated content that would be material to a consumer's decision — such as a founder letter, an expert opinion, or personalized outreach implying individual human judgment. A blog post attributed to a named human author but generated by an LLM, or a CEO voice email campaign written by AI, deceives consumers about the source of the advice they are relying on. Where authorship is a credibility signal, the FTC Act §5 deception standard applies to AI authorship just as it does to fabricated human authorship.

Add an AI-assistance disclosure to any content where authorship affects credibility, and restructure AI-written email campaigns to avoid false first-person voice.

function BlogPostMeta({ post }: { post: Post }) {
  return (
    <div className="text-sm text-gray-600">
      <span>By {post.author}</span>
      {post.aiAssisted && (
        <span className="ml-3 bg-gray-100 px-2 py-0.5 rounded text-xs">
          Written with AI assistance
        </span>
      )}
      <span className="ml-3">{formatDate(post.publishedAt)}</span>
    </div>
  )
}

// AI-generated mass email — avoid false individual voice
// AVOID: "I personally wanted to reach out to you, [name]..."
// (when the email is AI-generated to thousands of users)
//
// PREFER: "The [Product] team wanted to share..." or a clearly
//         branded automated format that does not imply individual
//         human authorship

The test: would a reasonable consumer make a different decision if they knew the content was AI-generated? If yes, disclose it. Commodity copy — FAQs, error messages, standard product descriptions — does not require disclosure where authorship is not a credibility factor.

• ID: ftc-consumer-protection.ai-decisions.ai-content-disclosed

• Severity: low

• What to look for: Count all relevant instances and enumerate each. Look for LLM API usage in content generation pipelines that produce consumer-facing text. Specifically: (1) blog posts generated by AI and published without disclosure; (2) product descriptions, help articles, or comparison pages drafted by AI and presented as expert human writing; (3) email campaigns generated by AI in a personalized voice that implies direct human authorship ("I personally want to share..."); (4) AI-written social proof content. The FTC's evolving guidance on AI distinguishes between AI as a writing aid (where disclosure is not always required) and AI-generated content that would be material to a consumer's decision — such as an AI-written expert opinion, an AI-impersonated founder letter, or AI-generated personalization that appears to be human judgment. Check: does the AI-generated content appear in a context where a consumer would care that it was AI-generated rather than written by a human expert or the actual person implied?

• Pass criteria: AI-generated content that could be material to a consumer's decision (expert reviews, founder letters, personalized recommendations framed as human judgment) is either disclosed or written in a way that does not imply human authorship. At least 1 implementation must be verified. Blog posts generated primarily by AI include a disclosure ("Written with AI assistance" or "AI-generated") if they are presented as expert human opinion.

• Fail criteria: AI-generated blog posts are attributed to a named human author without disclosure. AI-drafted email campaigns use first-person language implying a human wrote the specific message to that user. AI-generated "expert opinions" appear without disclosure in a context where authorship matters.

• Skip (N/A) when: The application uses no AI to generate static or asynchronous consumer-facing content (blog posts, product descriptions, email campaigns, landing page copy). AI chatbots and conversational interfaces are explicitly excluded from this check — they are covered by no-deceptive-ai-personas. This check concerns AI authorship of content artifacts, not AI-powered conversations.

• Detail on fail: Example: "Blog posts generated by OpenAI API are published under a named author's byline with no AI assistance disclosure." or "Monthly email campaign uses GPT-4 to write personalized outreach in the CEO's voice ('Hi, I wanted to personally reach out...') with no disclosure that the message is AI-generated." or "Product comparison page claims to be 'written by our expert team' but is generated by LLM with no human review or disclosure."

• Remediation: Add disclosure to AI-generated content where authorship is material:

  // Blog post with AI assistance disclosure
  function BlogPostMeta({ post }: { post: Post }) {
    return (
      <div className="text-sm text-gray-600">
        <span>By {post.author}</span>
        {post.aiAssisted && (
          <span className="ml-3 bg-gray-100 px-2 py-0.5 rounded text-xs">
            Written with AI assistance
          </span>
        )}
        <span className="ml-3">{formatDate(post.publishedAt)}</span>
      </div>
    )
  }
  
  // For AI-generated emails — use a clear sender voice
  // AVOID: "I personally wanted to reach out to you, [name]..."
  // (when the email is AI-generated to thousands of users)
  //
  // PREFER: "The [Product] team wanted to share..."
  //         or use a clearly branded automated email format
  //         that does not imply individual human authorship
  

  The test: would a reasonable consumer make a different decision if they knew the content was AI-generated? If yes, disclose it. If the content is commodity copy (standard product description, error messages, FAQs) where authorship is not material, disclosure is not required.