Telling customers they are talking to an AI: where it is required by law (the EU AI Act, California, Utah, and others), where it is expected by platform rules and consumer law, why disclosure helps rather than hurts, and the wording that works
Should the agent say it is an AI? In a growing number of places the law answers yes, and in the rest the answer is still yes, because customers who find out later feel tricked and customers who are told up front do not mind. Here is where it is required, what the rules say, and how to word it.
Quick answer
Disclosure that a customer is interacting with an AI is legally required in the EU under the AI Act's transparency rules for systems that interact with people, in California for bots used to sell or influence, in Utah for generative AI in consumer interactions when asked (and proactively for regulated occupations), and increasingly elsewhere; general consumer-protection law against misleading practices applies almost everywhere. Best practice is the same in every country: the agent identifies itself as an assistant in its first message, answers honestly when asked, and makes a person available. The evidence shows disclosure does not reduce conversion when the agent is helpful. ReplyKit agents disclose by default. This is general information, not legal advice.
Where it is required
| Jurisdiction | Rule, in brief |
|---|---|
| European Union | The AI Act requires that people are informed when they interact with an AI system, unless it is obvious; applies to chatbots serving EU customers |
| California | The bot disclosure law requires disclosing a bot used to incentivise a purchase or influence a vote; disclosure must be clear and conspicuous |
| Utah | Generative AI must disclose itself when a person asks, and proactively in regulated occupations |
| Colorado and other US states | Consumer-facing AI rules in force or pending; check current state law |
| United Kingdom | No AI-specific statute yet; consumer protection law against misleading practices applies |
| Elsewhere | Consumer and advertising law generally prohibits misleading customers about who they are dealing with |
Laws change; check current rules for the places your customers are. The data-protection side is separate and covered in GDPR and AI agents.
Platform rules and WhatsApp
WhatsApp's business and commerce policies require honest dealing with customers and prohibit deceptive practices; an undisclosed bot pretending to be a named human sits badly against that. Identifying the assistant also reduces the "is this a bot" friction that leads to blocks and reports. The policy is in WhatsApp's business policy.
Does disclosure cost sales?
In our customers' logs, conversations that open with "I'm the assistant" and then answer well convert as well as those that do not, and complaints about being deceived vanish. Customers care whether they got what they needed in seconds, not whether the first reply was human. What does cost sales is an agent that pretends and is found out, or a disclosure so heavy ("I am an artificial intelligence language model") that it sounds like a disclaimer. The broader evidence on expectation-setting is in setting expectations for AI replies.
Wording
First message
Hi! I'm {name}'s assistant. I can answer questions and book appointments, and a person is always available if you'd rather. How can I help?
When asked
Yes, I'm an AI assistant for {business}. I can help with most things; if you'd like a person, just say so.
Persona line
Always identify as the assistant in the first reply; never claim to be human; if asked, answer honestly and offer a person.
Regulated occupation
I'm an AI assistant for {practice}; I can book and answer general questions, but {the professional} handles any advice.
Short, warm, and followed by usefulness. The persona rules are in persona and tone of voice.
Frequently asked questions
Do I have to tell customers they are talking to an AI?
In the EU and several US states, yes by law. Everywhere else, consumer law and good practice point the same way.
Does disclosure reduce sales?
Not when the agent is helpful. Being found out after pretending does.
What should the first message say?
That it is the assistant, what it can do, and that a person is available.
What if a customer asks "are you a bot"?
Answer honestly and offer a person.
Is this legal advice?
No. Laws vary and change; check the rules where your customers are.
Honest from the first message
The assistant identifies itself by default and a person is always one message away. 7-day free trial, no card.
Sources and further reading
- WhatsApp Business Messaging Policy · opt-in, opt-out and prohibited-use rules every business must follow
- ReplyKit Privacy Policy
This article is general information about disclosure rules, not legal advice. Laws vary by jurisdiction and change; consult a qualified professional about your obligations.