Imagine: you write to customer support, have a lively conversation, feel genuinely attended to and then discover it was all an AI agent. Some customers react indifferently: “So what, my issue was resolved.” Others feel deceived. This reaction lies at the heart of one of the sharpest debates in CX: communication transparency should a company disclose that a customer is speaking with a bot? In this article, we examine the arguments from both sides and give a clear recommendation.
Whether to Disclose the Bot: Arguments For
Proponents of mandatory disclosure rely on several strong arguments:
Trust as a long-term investment. A customer who was deceived once will not return. Research shows that 65% of consumers who discover they were speaking with a bot presented as a human report reduced customer trust toward that brand. Conversely, companies that honestly disclose AI from the first seconds of interaction receive higher satisfaction scores.
Regulatory pressure. By the end of 2023, the EU had agreed on the text, and in 2024 it officially adopted the AI Act, one of whose provisions explicitly prohibits AI systems from posing as humans without clearly warning the user. Similar AI usage rules are already in development or in force in California, Canada, and several Asian countries. For companies entering international markets, compliance is a matter not only of ethics but of legal risk.
Expectation management. When customers know they are talking to an AI, they subconsciously adjust their expectations: they formulate requests more clearly, do not expect warm personal greetings, and are less frustrated by limited empathy in responses.
Arguments Against: Where the Logic Breaks Down
Companies that avoid disclosure often cite the following reasoning:
- “Customers do not want to talk to a bot” and they fear that disclosure will reduce engagement.
- “Our agent is so high-quality that it makes no difference” and they consider disclosure unnecessary.
- “It will lower conversion” prioritizing short-term metrics over trust.
The problem with these arguments is that they are oriented toward short-term results. Chatbot ethics is not merely philosophy. It is practical risk management: one viral scandal about “a bot pretending to be human” can cost far more than a short-term conversion boost.
The Middle Ground: How to Disclose AI Correctly
Leading company practice shows: disclosure does not need to be alarming or off-putting. It should be natural and can even become a competitive advantage.
Examples of correct framing:
- “Hello! I’m Aisha, the AI assistant for Company N. I’ll answer your questions instantly — and if needed, I’ll connect you to a live specialist right away.”
- “Our smart assistant is here to help you right now — no queue, any time of day.”
Note: in the first example, the agent has a name and personality — but honestly identifies as AI. This does not reduce customer trust — on the contrary, it strengthens it through predictability.
Three mandatory moments for disclosure:
- At the start of first contact — always, when the customer initiates the conversation.
- When transferring from AI to an operator — “I am now connecting you to a live specialist.”
- If the customer directly asks: “Are you a human or a bot?” — the AI must answer honestly.
AI Usage Rules and Brand Reputation
Commitment to AI usage rules is a litmus test for brand reputation in the long run. Companies that choose transparency today are building an asset: an audience that understands how and why AI is used, and trusts that choice.
Notably, the highest CSAT scores in AI support do not go to systems that best “mimic” humans, but to those that honestly and effectively resolve problems. What matters to the customer is that the issue is solved — not that the respondent has a human voice.
Communication transparency and high-quality AI responses are not a compromise but a synergy. Intelswift allows you to configure the agent’s personality, name, ToV, and identification rules in line with your company’s brand guidelines. Want to see how it looks in practice — request a demo.
