How to Disclose AI Use Without Losing Customers: Scripts That Convert

The fear is always the same: "if I say it's an AI, they'll hang up on me." The reality is the exact opposite. The question isn't whether to disclose, but how to disclose AI use without losing customers — and that's where nearly the whole game is won. What makes people hang up isn't the word "AI"; it's the feeling of being deceived, or a phone menu that leads nowhere. A well-written disclosure — clear, short, with an exit to a person — builds trust, meets the rules, and keeps the conversation alive. In this guide we give you real scripts you can copy, for voice, WhatsApp, web chat and email, and we show you exactly where and when to say them.

Why disclosing AI use works in your favor

The EU's AI Act makes disclosing that an AI is answering more than a courtesy: its transparency duty requires the provider to design the system so a person knows they're interacting with an AI, unless that's already obvious (Regulation (EU) 2024/1689 —the AI Act—, art. 50.1). The applicability part is what a US business needs to hear: if you take calls from or handle data of people in the EU, that duty applies to you directly. If you operate purely in the US, no EU regulator is at your door — but disclosing AI is already what the FTC and a growing set of state laws expect, and the AI Act is simply the strictest baseline you can build to (art. 113). On the data side, GDPR (which kicks in when you serve EU residents) grants a right to human intervention only for solely automated decisions with legal or similarly significant effects — something a booking bot normally doesn't make (art. 22.3 GDPR). (US privacy laws like CCPA/CPRA don't grant an equivalent human-intervention right, so this is an EU-law point, not a US one.) Even so, always offering to hand off to a person is simply the best trust practice. The temptation is to treat all this as a burden. That's the mistake.

Think about it from the other end of the line. When someone tells you up front what's going on, you lower your guard. When you find out halfway through — because the voice "catches" or the chat repeats itself — you feel duped and you shut down. Disclosure doesn't cost you anything: it removes the one truly awkward moment there is — the moment of suspicion.

If you want the full legal framework behind this disclosure, we lay it out in whether it's legal to use an AI voice agent or chatbot for customer service and in the checklist of the AI Act's key obligations for your business.

Anatomy of a good AI disclosure: the 3 pieces

Call it an AI disclosure, a transparency clause, or a virtual-assistant introduction: the name doesn't matter, what matters is that it has three pieces, not one more.

Those three pieces solve compliance and conversion at once. "Clear" satisfies the AI Act's transparency duty (art. 50.1). "With an exit" goes beyond what's strictly required: GDPR guarantees human intervention only when there's a solely automated decision with legal or significant effects (art. 22.3 GDPR), not as a general right to talk to a human in any conversation; offering that exit is, above all, an excellent trust practice, as we explain in what GDPR Article 22 actually covers. And "short" is what keeps your customer from getting impatient before they even reach the reason for their call.

Tone is everything. The difference between "please be advised that you are interacting with an artificial intelligence system" and "I'm the clinic's virtual assistant, happy to help" is the same legal sentence told two ways. One sounds like a courtroom. The other, like a front desk.

AI disclosure scripts: the ones that convert vs. the ones that scare people off

Here's the practical heart of it: AI disclosure examples ready to copy. Same message, two ways to say it: the one that ends the call and the one that keeps it going.

See the pattern? The disclosure that scares people off is defensive: it warns of a risk and asks permission. The one that converts is hospitable: it introduces itself, offers help, and leaves the door open to a person without making it feel like a penalty. Customers read the second version as service, not fine print.

That line takes about four seconds, satisfies the AI Act, offers the human exit, and is already asking why they're calling. There's nothing to "sit through": the conversation moves forward.

Where to put your AI disclosure: voice, WhatsApp, web chat and email

The principle is simple: once, at the start of every conversation, whatever the channel. Repeating it in every message gets tiring; hiding it does too. Change the format by channel, but never the substance: from the first line, the customer knows an AI assistant is helping them and that they can ask for a person.

1. 1

   Voice call

2. 2

   WhatsApp, Instagram and Messenger

3. 3

   Web chat on your site

4. 4

   Email

One detail makes all the difference: the disclosure has to come with a useful answer right away. Declaring that you're an AI is worthless if you then take thirty seconds to understand the question. Transparency rests on usefulness. That's why disclosure works best in a system that picks up the phone in ~1 second and answers your messages right away, not one that keeps people waiting.

How Totem says it naturally on every channel

At Totem, the AI disclosure isn't a patch you stick on top: it comes built into every channel — voice, WhatsApp, Instagram, Messenger, web chat and email — with the wording you approve before it goes live. The AI answers first, introduces itself in your brand voice, and at any moment a human on your team picks up the conversation in one click, with the full context already in front of them. That's exactly the "hand off to a person" the rules point toward, made real without friction. We cover it in depth in how Totem builds in AI disclosure, consent, EU data residency and the handoff and in how the AI hands the conversation to a person with full context.

And the underlying question — do people prefer to talk to a person or an AI? — has a data-backed answer that might surprise you; we dig into it in what the data says about the transparent hybrid model.

The conclusion is the one that opened this guide, but now with scripts in hand: AI disclosure isn't the enemy of your conversion rate. The enemy is saying it badly — cold, late, or hidden. Say it clearly, say it briefly, offer the person, and answer fast. Transparency, well written, sells.

Official sources

• AI Act (Regulation (EU) 2024/1689), art. 50 (transparency) and art. 113 (application timeline) — EUR-Lex
• GDPR (Regulation (EU) 2016/679), art. 22 (automated decisions) — EUR-Lex