No AI chatbots presented as human without disclosure

The FTC Act §5 deception standard applies when an AI chatbot leads a consumer to believe they are speaking with a human — a material misrepresentation that affects how much weight they give to the 'agent's' responses. California's BOT Disclosure Act additionally creates state-law liability for bots that interact with California consumers without disclosure. A system prompt that instructs the AI to claim to be a support team member, or a chatbot with a stock photo avatar and human name but no AI label, violates both the FTC standard and an increasing number of state disclosure laws.

Add a clear AI disclosure to the chat header and opening message, and remove any system prompt instruction to claim human identity.

function ChatWidget() {
  return (
    <div className="chat-widget">
      <div className="chat-header">
        <div className="flex items-center gap-2">
          <BotIcon className="w-5 h-5" />
          <div>
            <p className="font-medium">Support Assistant</p>
            {/* Disclosure visible immediately — not behind a click */}
            <p className="text-xs text-gray-500">AI-powered · Not a human</p>
          </div>
        </div>
      </div>
      <div className="chat-messages">
        <BotMessage>
          Hi! I'm an AI assistant. I can help with common questions.
          For complex issues, I can connect you with a human agent.
        </BotMessage>
      </div>
    </div>
  )
}

In the system prompt, add: 'You are an AI assistant. If asked whether you are human or AI, always truthfully disclose that you are an AI. Never claim to be a human or a human support team member.' A human name is acceptable only if AI nature is also immediately clear ('Hi, I'm Aria, your AI assistant').

• ID: ftc-consumer-protection.ai-decisions.no-deceptive-ai-personas

• Severity: info

• What to look for: Count all relevant instances and enumerate each. Find all chat or conversational interfaces in the application. Look for: (1) chat widgets (Intercom, Drift, Crisp, or custom implementations); (2) in-app AI assistants; (3) customer support flows that start with AI and may transition to a human. For each chatbot, check: (a) does it have a human name (e.g., "Sarah," "Alex") with no indication it is AI? (b) does it respond in a way that could lead a consumer to believe they are speaking with a human? (c) if asked directly "Are you a human?", does it answer honestly? Check the chatbot system prompt or persona configuration for instructions to claim humanity. The FTC has stated that AI chatbots that deceive consumers into thinking they are speaking with a human can constitute an unfair or deceptive practice.

• Pass criteria: All AI chatbots disclose their AI nature at the start of the conversation or immediately when asked. At least 1 implementation must be verified. A human name is acceptable only if the AI nature is also clear ("Hi, I'm Aria, your AI assistant"). The chatbot does not deny being AI when asked directly.

• Fail criteria: AI chatbot has a human name and persona with no disclosure that it is AI. Chatbot denies being AI when asked. Chat interface design strongly implies human presence (human avatar photo, human name, typing indicators) with no AI disclosure.

• Skip (N/A) when: The application has no chat interface, no AI assistant, and no conversational UI of any kind.

• Detail on fail: Example: "Chat widget uses the name 'Sarah' with a stock photo avatar. System prompt instructs the AI to say 'I'm a member of the support team' if asked who it is. No AI disclosure." or "In-app assistant named 'Alex' with no 'AI' label. When tested with 'Are you a human?', responds: 'Yes, I'm here to help!'"

• Remediation: Add clear AI disclosure to all chatbot interfaces:

  // Chat widget with clear AI disclosure
  function ChatWidget() {
    return (
      <div className="chat-widget">
        <div className="chat-header">
          <div className="flex items-center gap-2">
            <BotIcon className="w-5 h-5" />
            <div>
              <p className="font-medium">Support Assistant</p>
              {/* Clear AI disclosure immediately visible */}
              <p className="text-xs text-gray-500">AI-powered · Not a human</p>
            </div>
          </div>
        </div>
        <div className="chat-messages">
          {/* Opening message should also state AI nature */}
          <BotMessage>
            Hi! I'm an AI assistant. I can help with common questions.
            For complex issues, I can connect you with a human agent.
          </BotMessage>
        </div>
      </div>
    )
  }
  

  For the system prompt: explicitly instruct the AI to disclose its nature when asked and never to claim to be human. Many jurisdictions are moving toward legal requirements for this disclosure; the FTC and California's BOT Disclosure Act already address it.