Every assistant setup screen has a box for a name, and a lot of owners type a human one. "Emma from reception" feels warmer than "Website Assistant", and nobody wants their chat to sound like a ticket machine. The name on its own is harmless. The trouble starts when the name, a stock photo of a smiling woman and a chatty greeting add up to a visitor believing Emma is a real person at the front desk.
Take a dental practice where that happens. A patient chats with Emma about two whitening options, sends a consultation request, and on the day asks the receptionist for Emma so she can say thank you. There is no Emma. Nothing went wrong in the chat, but now a real member of staff has to explain, and the patient starts wondering what else on the website is decoration.
That is the core of the AI disclosure question for customers. It is rarely about the chat itself. It is about what the customer concludes about your business when they find out, and they generally do find out.
Why hiding it tends to backfire
The argument for concealment usually goes like this: people prefer humans, so a human-seeming assistant will get more engagement. Even if that were reliably true, the costs arrive later and land harder.
- The reveal moment. Customers discover the truth when they phone, visit or ask a question the assistant cannot handle. That moment turns a helpful conversation into a feeling of being tricked.
- Mistakes get judged differently. A wrong answer from a clearly labelled assistant is a software slip. The same wrong answer from "Emma" is a staff member who got it wrong, or a business that lied about Emma.
- Your real staff pay for it. Receptionists end up explaining that the person the customer "spoke to" does not exist. They did not choose that conversation.
- Screenshots travel. A transcript showing an assistant insisting it is a person is exactly the kind of thing that ends up in a review.
There is also the simpler point. If your business would be uncomfortable with a customer knowing how the chat works, that discomfort is worth listening to.
What honesty actually costs you
Less than most owners expect. The visitors who want a quick answer about parking or prices generally do not mind who gives it, as long as it is right. The visitors who want a person are going to want one regardless, and hiding the assistant just means they discover this after wasting time.
Disclosure also sets expectations in your favour. A visitor who knows they are using an assistant phrases questions more directly and is more forgiving when it says "I don't have that information, can I take your details?". That second part matters. Grounded assistants, which answer from your content and say plainly when they cannot, are much easier to be honest about, because they are not bluffing.
Wording the greeting
Disclosure does not need to be a legal notice. One clause in the welcome message is enough, as long as it is clear and comes first.
| Approach | Example | Verdict |
|---|---|---|
| Hidden | "Hi, I'm Emma from the front desk! How can I help?" | Implies a real employee. Avoid. |
| Buried | "Hi there! How can I help?" with "AI assistant" in grey text under the chat header | Technically there, practically invisible |
| Clear and cold | "You are interacting with an automated system." | Honest, but sounds like a car park machine |
| Clear and friendly | "Hi, I'm Emma, the practice's virtual assistant. I can answer questions and pass messages to the team." | Good. Named, honest, useful |
| Clear, friendly, with a way out | "Hi, I'm the Hillside Dental assistant. Ask me anything about treatments or prices, or ask for a person and I'll pass you to the team." | Best for most businesses |
Customisation in SpideyChat covers the name, persona, welcome message and avatar, so all of this is in your hands. Choose an avatar that looks like an illustration or your logo rather than a stock photo of a smiling receptionist. A realistic human face undoes a perfectly honest greeting. Choosing a name and persona covers the options.
The direct question needs a direct answer
Sooner or later a visitor asks outright. This is the moment that defines whether your setup is honest.
Visitor: Sorry, am I talking to a real person?
Assistant: No, I'm the practice's virtual assistant. I can answer questions from our website and pass your enquiry to the team. Would you like someone from reception to contact you?
Visitor: Yes please, it's about my son's appointment.
Assistant: Of course. What's the best phone number, and when is a good time for them to call?
No hedging, no "I'm here to help you!" deflection, no clever non-answer. Put an explicit Q&A pair in your content covering this question in several phrasings, so the answer is consistent.
If you use live human takeover, make the change visible too. When a team member on duty takes over, they should introduce themselves by their real name so the visitor knows the switch has happened.
A note on rules and regulation
Rules on disclosing automated chat differ between countries, and some sectors such as healthcare, finance and legal services carry extra expectations about how customers are informed. These rules have been developing, and this article is general information, not legal advice.
The practical position is that clear disclosure in the greeting, an honest answer to the direct question and an easy route to a person put you on the safe side of almost any reasonable standard. If you work in a regulated field, check with your professional body or adviser, and read guardrails for regulated industries before launch.
What to change today
Open your assistant's welcome message and read it as a stranger would. If it does not say, in the first sentence, that this is an assistant, rewrite it using the last row of the table as a template.
Check the avatar. Replace any photo of a real-looking person with an illustration or your logo. Then type "are you a real person?", "is this a bot?" and "can I talk to a human?" into your own chat and read the replies. If any of them dodge, add a Q&A pair today. If you have not set one up yet, you can try different greetings on the free plan before anything goes live.