Skip to content
AI-SNS.IO — CHAT SUPPORTAI-SNS.IO — CHAT SUPPORT
Security & Compliance

Disclosing That Customers Are Talking to AI

AI SNS CHAT EditorialPublished 7 min read

How to tell customers an AI is replying on web chat, WhatsApp and Messenger: wording, timing, handoff promises and a practical setup checklist.

Customers increasingly ask, sometimes bluntly, "Am I talking to a real person?" If your answer is vague, late or dodged, you lose trust faster than any wrong reply could. This article shows how to tell people an AI is responding, where to say it, what wording works, and how to keep the promise you make about reaching a human.

Why disclosure is now part of good support

Telling customers they are talking to AI is no longer just a courtesy. Several jurisdictions have transparency rules for automated conversations, including the EU's AI Act and California's bot disclosure law. The details and scope differ by place and by use case, so confirm your obligations with your own legal counsel. Even where no rule applies, there are practical reasons to be upfront.

  • Expectations match reality. People who know it is an AI write shorter, clearer questions and forgive a "let me connect you" more easily.
  • Trust survives mistakes. A customer who was told at the start treats an error as a limitation. A customer who finds out later treats it as deception.
  • Escalation is cleaner. When the AI is named as the first line, asking for a person feels natural rather than like a complaint.
  • Sensitive topics get the right care. Billing disputes, health questions and complaints are where people most want to know who is answering.

Disclosure is not an apology. You are not saying "sorry, it's only a bot." You are saying what the service is, what it is good at, and how to get more.

What a good disclosure contains

A strong disclosure is short and does four jobs at once:

  1. States that an AI is replying. Use the plain words "AI assistant" or "automated assistant."
  2. Says what it can help with. One concrete line, such as orders, bookings or product questions.
  3. Sets a limit. Mention that it can make mistakes on important matters, or that it will not handle certain things.
  4. Offers a human route. Say how to reach a person, without inventing a promise you cannot keep.

Wording that works

Compare that with the versions that tend to cause trouble:

Vague or risky
Clear and honest
"Hi, I'm Emma!" with a human-looking photo and no mention of AI
"Hi, I'm Emma, the shop's AI assistant."
"Our team is here to help"
"I'm an AI. A team member can take over if you ask."
"Ask me anything"
"I can help with orders, delivery and returns."
"A human will reply in 5 minutes" (no rota behind it)
"I'll pass this to the team and tell you what happens next."
Disclosure buried in the terms page
Disclosure in the first message the customer sees
Disclosure wording: vague versus clearIllustrative example

A friendly name is fine. A human-sounding name with a stock photo and no label is where disclosure breaks down.

Where and when to say it

The right moment is before the customer has invested effort, which means the first message, not the footer.

Website chat

Put the disclosure in the welcome message and, if your widget supports it, in the header or launcher text as well. The welcome message is the one thing every visitor reads, so it carries the most weight. If your site serves several languages, make sure the disclosure appears in the visitor's language rather than only in English.

WhatsApp

WhatsApp conversations often start with the customer sending a message, so the first reply from your agent should open with the disclosure. Keep it to one or two lines. Customers read WhatsApp on small screens and expect brevity. Your business profile description is a good second place to state that replies are handled by an AI assistant with human backup.

Messenger and Instagram

Use the same structure as on WhatsApp. If a customer arrives from an ad or a comment, they may not have any context about who they are messaging, which makes the first-reply disclosure even more important.

Voice

Phone is the channel where people most assume a human. Open the call by saying it is an AI assistant, before the customer starts explaining a problem. If you use outbound calling, the disclosure belongs in the opening sentence, ahead of the reason for the call.

Keep the human-handoff promise honest

Disclosure creates an implied promise: "if you need a person, you can get one." Breaking that promise is worse than never disclosing.

  • Do not state a wait time you cannot guarantee. "A specialist will reply within 5 minutes" only works if someone is actually on shift.
  • Tell people what happens next. "I've passed this to the team. They'll reply here during opening hours" is truthful and sets expectations.
  • Make the trigger obvious. A visible phrase such as "type human" or a button works better than hoping customers guess.
  • Hand over the context. Customers should not have to repeat themselves after the transfer.

Here is the shape of a healthy flow:

  1. 01AI introduces itself in the first message
  2. 02Customer asks a question or requests a person
  3. 03AI answers, or hands off with a clear next step
  4. 04Team member takes over with the history
Disclosure to handoffIllustrative example

Tone: honest without being cold

Some teams worry that saying "AI" will make customers leave. In practice the risk is more often in how it is said. A few habits help:

  • Be matter-of-fact. One sentence, no drama, no long disclaimers.
  • Avoid pretending to feelings or experiences. An AI should not claim to have "personally tried" a product.
  • Answer the question directly when asked. If a customer asks "are you a bot?", the agent should say yes, plainly, every time, even mid-conversation and even if the persona is playful.
  • Do not overpromise accuracy. For anything involving money, health, legal matters or safety, say that a person should confirm.
  • Keep your brand voice. Disclosure and personality are compatible. "I'm an AI, but I'll still try to make returns painless" is fine.

Common mistakes to avoid

  • Disclosing only in the terms or privacy page. Few customers read them, so they do not count as real notice for the person in the chat.
  • Using a human photo and a human name without any label. Customers will feel misled once they notice.
  • Letting the agent deny being an AI. Add an instruction to the persona so it always answers honestly when asked.
  • Disclosing once on the website but not on messaging apps. Each channel is its own first impression.
  • Forgetting translated versions. A disclosure that exists only in English is missing for everyone else.
  • Mixing up automation and people in handoffs. After a human takes over, say so. Customers should know when the speaker changes.

A short checklist before you go live

  1. The first message on every channel says an AI is replying.
  2. The wording is in each language you serve.
  3. The persona instructions tell the agent to confirm it is an AI whenever asked.
  4. There is a visible way to ask for a person.
  5. The handoff message describes the next step without inventing a time.
  6. You have tried it as a customer on web chat, WhatsApp and Messenger.
  7. Someone reviews real conversations regularly to see how disclosure lands.

Doing this in AI SNS CHAT

AI SNS CHAT gives you several places to build disclosure in, without adding anything that is not already in the product. Features below are limited to what the platform offers today.

Welcome message. Write your disclosure into the agent's welcome message. It is automatically translated into every language the widget speaks, so one well-written sentence reaches visitors in their own language. Check the translations read naturally for your market.

Persona and guardrails. In the agent's persona instructions, state that it is an AI assistant, that it must say so honestly whenever asked, and what topics it should send to a person. You can start from one of the starter personas, use the prompt generator or prompt library, and then edit the text to match your tone.

Test before launch. Use Playground to try conversations such as "are you a human?" and "I want to talk to someone." Playground also supports side-by-side model comparison, which helps if you want to see whether different models phrase the disclosure differently.

Channels. The same agent can be deployed to the chat widget, WhatsApp, Messenger, Instagram and many other channels listed under Channels in the sidebar. Because the first impression differs on each, open each channel and read the first exchange yourself.

Handoff. When a conversation goes to a person, the "what happens next" line is added by the server from your rota, and it never invents a wait time. Set shifts under Helpdesk → Scheduling (Standard plan or higher) and handoffs are auto-assigned to whoever is on shift. Alerts can go to Slack, Teams, LINE WORKS, WhatsApp, email or a webhook. You can also take over a chat directly, which pauses the bot, from Activity → Chat logs (or Inbox for email).

Review. Chat logs let you filter, flag, escalate and rate conversations with thumbs up or down. Make it a routine to read a sample and check that disclosure and handoffs behave as written.

Outbound messaging. Outbound campaigns (Standard plan or higher) only reach people who messaged first, and every message carries a STOP opt-out. Add a short AI disclosure to your templates as well.

On privacy, AI SNS CHAT describes its practices as designed with reference to GDPR and APPI. It does not claim SOC 2, ISO 27001 or HIPAA certification. Details are on the security page. For plans and limits, see pricing, and for the channel list, see connections and features.

Frequently asked questions

Do I have to say "AI" explicitly, or is "virtual assistant" enough?

"Virtual assistant" alone can be read as a human working remotely. Pairing it with "AI" or "automated" removes the doubt. Whether specific wording is legally required depends on your jurisdiction, so check with your counsel.

Will disclosing AI reduce how many customers use the chat?

There is no universal answer, and you should measure it for your own audience. A clear, friendly first message that explains what the assistant can do and how to reach a person tends to feel more helpful than a mysterious one. Compare chat logs and ratings before and after changing your wording.

Should the agent keep reminding people it is an AI?

Not in every message. Disclose at the start, then answer honestly whenever the customer asks, and mention it again at handoff points or when the topic is sensitive.

Can I give my AI agent a human-style name?

Yes, as long as the label is clear. A name like "Emma, the shop's AI assistant" gives personality without misleading anyone. Avoid pairing a human name and photo with silence about what it is.

How do I check that my disclosure actually shows up on every channel?

Open each channel as a customer would, start a fresh conversation and read the first reply. Then ask directly "are you a person?" and ask for a human. Repeat this after any change to the welcome message or persona.

#AI disclosure#transparency#customer support#compliance#chatbots

Put this into practice with AI SNS CHAT

Build an AI agent trained on your own content and connect it to your website, LINE, WhatsApp and more. Free to start, no credit card required.