Do You Have to Tell Customers They're Chatting With AI?
More small businesses are putting AI assistants on their websites to answer common questions, collect leads and cover the hours nobody is at a desk. That raises a fair question: if a visitor is chatting with software, do you have to tell them?
The short answer is that in a growing number of places, yes, at least in some situations. And even where no law requires it, telling people up front is usually the smarter move. This guide walks through the main rules in plain English and then the practical side: how to disclose AI chat in a way that feels natural and keeps customers' trust.
This is general information, not legal advice. Laws change, they apply differently depending on your business and your customers, and the details matter. Check the rules where you operate and where your customers are, and talk to a lawyer if you are unsure.
The rules in brief
There is no single worldwide rule for AI chat. Instead there is a patchwork, and it is still being filled in. Here are the ones small businesses ask about most.
European Union: the AI Act (Article 50)
The EU AI Act includes transparency rules for AI systems that interact directly with people. In general terms, AI systems meant to talk with people must be built so that people are told they are dealing with an AI, unless that is already obvious to a reasonably well-informed person from the context. The information has to be given clearly, at the latest at the first interaction.
These transparency obligations began to apply on 2 August 2026. A 2026 amendment package (often called the Digital Omnibus) pushed back several other parts of the AI Act, mainly the rules for high-risk systems, but the chatbot disclosure duty was not among the delayed items.
Formally, this duty sits mostly with the provider of the AI system (the company that builds it), not the small business that adds a chat widget to its site. In practice, if you serve customers in the EU, you should choose a tool that discloses AI clearly and not switch that disclosure off or hide it.
California: the bot disclosure law (SB 1001)
California's bot law, passed as SB 1001 and in effect since July 2019, makes it unlawful to use a bot to communicate with people in California online with the intent to mislead them about its artificial identity in order to influence a purchase or sale of goods or services (or a vote in an election). The law gives a clear way to stay on the right side of it: disclose that it is a bot, in a way that is clear, conspicuous and reasonably designed to inform the person.
A later bill that would have expanded California's bot disclosure rules (AB 410, 2025) stalled in the state Senate, so the 2019 law is the main one to know.
Utah: disclosure when asked, and more for licensed professions
Utah was early to pass a broad AI consumer law, and it has been amended since. Under the current version, as generally described, a business using generative AI to interact with consumers must disclose that the person is talking to AI if the consumer clearly asks. Stricter rules apply to licensed, regulated professions in higher-risk interactions (for example, sensitive personal, financial or health matters), where the disclosure has to be made up front.
Maine: chatbot disclosure in trade and commerce
Maine passed a chatbot disclosure law in 2025. In broad terms, if a business uses an AI chatbot or similar technology with consumers in a way that could mislead a reasonable consumer into believing they are talking with a human, it must tell the consumer clearly and conspicuously that they are not. A violation is treated as an unfair trade practice under Maine law.
The common thread
Notice what these rules have in common. None of them bans AI chat. They target deception: letting people believe a bot is a person. And the safest way to comply with all of them at once is the same simple habit: tell people clearly, at the start, that they are chatting with AI.
Other states have passed or are considering their own chatbot rules, some aimed at specific uses such as companion or mental health bots. That list keeps growing, which is another reason to treat clear disclosure as your default rather than tracking each rule one by one.
Why disclose even when you do not have to
Even if none of these laws reaches your business today, there are good reasons to be upfront.
- People find out anyway. AI answers have a recognizable rhythm. A customer who works out mid-conversation that the friendly "person" was a bot tends to feel tricked, and that feeling lands on your brand, not the software.
- It sets the right expectations. When visitors know they are talking to an assistant, they ask simpler, clearer questions and are less frustrated when it hands off to a person.
- It protects you as rules spread. Disclosure rules have been moving in one direction. Building the habit now means you will not have to rework your chat later.
- Honesty is part of good service. Customers are generally fine with AI that helps them quickly. What bothers them is not being told.
If you are still deciding between a bot, a human, or a mix, our comparison of AI chatbots and live chat covers the trade-offs.
How to disclose AI chat well
Good disclosure is not a legal wall of text. It is a few small design choices that make the situation obvious without getting in the way.
1. Say it in the first message
The first thing the assistant says should make clear it is AI. Keep it short and friendly, for example:
- "Hi, I'm Ava, an AI assistant for Bright Smile Dental. I can help with hours, insurance and booking. Want a person instead? Just ask."
- "Hello! I'm the AI assistant here. I can answer quick questions about our services, or connect you with our team."
Your greeting does a lot of work in chat, so it is worth getting right. Here are more live chat greeting examples you can adapt.
2. Keep a visible AI label on every message
A first message scrolls away. A small "AI" badge or label next to each assistant message keeps the disclosure in view for the whole conversation, including for someone who joins late or scrolls back later. It also helps when a human team member takes over, because the difference between AI and human replies stays clear.
3. Make it easy to reach a person
A visible "Talk to a person" option (or a clear way to leave a message for your team) shows respect for the visitor's choice. Some people simply prefer a human, and some questions, like refunds, complaints or anything sensitive, should go to one. If nobody is available, say when someone will reply instead of leaving them guessing.
4. Never deny being AI
This is the line that matters most. If a visitor asks "Am I talking to a real person?", the assistant should answer honestly, every time. Instructions that tell a bot to dodge or claim to be human are exactly what these laws are aimed at, and a single screenshot of a bot lying can undo a lot of goodwill.
5. A human name is fine, a human pretense is not
Giving your assistant a name and an avatar can make it feel more approachable. Just pair that with the label and the first-message disclosure so the personality never becomes a disguise.
How Chattiv handles it
We built disclosure into Chattiv so you do not have to remember it. Chattiv is human-first: your team answers chats, and the AI assistant, Ava, is optional. When Ava does reply, every one of her messages carries an AI badge, and her first message says she is an AI assistant. You can rename her and change her avatar, but the badge stays. Ava never claims to be human, and a "Talk to a person" option stays visible so visitors can ask for your team at any time. You can see how it works on the Ava page.
The bottom line
Do you have to tell customers they are chatting with AI? Depending on where you and your customers are, the law may already require it, especially in the EU, in California when a bot is used to sell, and in Maine. Utah requires it when a customer asks, and sooner for some professions. Everywhere else, it is still the approach that keeps trust intact.
The practical version is simple: say it in the first message, keep an AI label visible, make a person easy to reach, and never let the bot pretend to be human. And because the rules keep changing, check the current requirements where you do business, or ask a lawyer, before you rely on any summary, including this one.
Frequently asked questions
Is it enough to say I use AI in my privacy policy?
Usually not. The rules that exist focus on telling people clearly at the moment they are chatting, for example the EU AI Act asks for the information at the latest at the first interaction, and Maine and California use a clear and conspicuous standard. A line buried in a policy page is unlikely to meet that. Put the notice in the chat itself.
Can my AI assistant have a human name and avatar?
Many businesses give their assistant a friendly name, and that is common practice. The key is that the name and picture should not be used to suggest a person is typing. Pair them with a visible AI label and a first message that says it is an AI assistant, and have it answer honestly whenever someone asks.
What if a person and the AI both answer in the same chat?
Make it clear which is which. Label each AI message, and when a team member joins, say so (for example, Sam from our team has joined the chat). That way the customer always knows who they are talking to, and nobody mistakes a human reply for the bot or the bot for a human.
Try Chattiv free
Live chat for your website and a team inbox on your phone, with Ava, an optional AI assistant that always says she is AI and hands off to you. Free plan, no card.
Start free