Should You Tell Customers They're Talking to AI? A Practical Disclosure Policy
TL;DR
If a customer could reasonably mistake AI for a human, or if AI is shaping a decision that affects them, tell them, plainly, early, and with a route to a real person. Disclosure done well builds trust rather than killing it, and a one-line policy beats a legal essay every time.
Should you tell customers they're talking to AI at all?
Yes, when there's a genuine chance they'd think they were dealing with a person and it matters to them. That's the whole test, really. You don't need a committee or a policy the length of a tax return. You need to ask one honest question: would this customer feel tricked if they found out later?
Most owners I talk to are nervous about this. They worry that the moment they admit AI is involved, the customer assumes the service is cheap, lazy, or that they've been fobbed off. I understand the fear. But I've watched it play out the other way far more often. The customer who feels conned, who spent ten minutes pouring their heart out to what they thought was a person, is the one who leaves a one-star review and never comes back.
AI disclosure to customers isn't a confession. It's a courtesy. And courtesy is good business.
When does disclosure actually matter?
Disclosure matters in two situations: when the customer is interacting with AI directly, and when AI is shaping a decision that affects them.
The first is obvious. A chatbot on your website, an AI voice answering the phone, an automated reply that sounds like it came from a colleague, if a person on the other end might assume they're speaking to a human, say so.
The second is quieter and easier to miss. If AI is helping decide someone's price, their eligibility for a service, their place in a queue, or whether their claim gets approved, that affects them even though they never "spoke" to the machine. People deserve to know when a decision about them was shaped by software, especially if they might want to question it.
Here's where it does not matter: AI working in the background that touches no customer-facing decision. If AI drafts your internal stock report, tidies your inbox, or summarises a meeting for your own use, there's nothing to disclose. Nobody needs a label on the tool you use to think. The line is about the customer's experience and the decisions that land on them, not about every place a model happens to run.
See where AI fits in your business. Free.
A 45-minute audit. We map the highest-value automations and what they're worth in time and money. No pitch, no pressure.
What does a good disclosure policy look like?
A good policy fits on one page and answers four questions: where, when, how, and who.
Where AI touches the customer, list the spots. Website chat. Phone line. Email replies. Any decision tool. If you can't list them, that's your first job.
When you tell them, at the first point of contact, not buried three clicks deep. Upfront beats clever. A customer who learns at the start that they're chatting with an assistant feels respected. One who works it out halfway through feels played.
How you say it, in plain words, the way you'd say it across a counter. Not "this interaction may be facilitated by automated systems." More like, "You're chatting with our AI assistant, ask for a person any time."
Who they can reach instead, always give an exit. The single biggest trust-builder isn't the disclosure itself, it's the easy route to a human when the customer wants one. AI should augment your team's reach, not wall customers off from it.
That's the whole policy. Where, when, how, who. You could write it this afternoon.
How do you word it without sounding like a robot?
Write it the way you'd say it out loud, then cut it in half. The fastest way to make a disclosure feel cold is to lawyer it up.
Compare these two. A bank-style version: "Please be advised that your enquiry may be processed using artificial intelligence technologies in accordance with our terms of service." Nobody reads that, and the ones who do feel like they've been handed a waiver.
Now a human version: "Hi, you're chatting with our AI assistant. I can sort most things quickly, and if you'd rather talk to one of the team, just say the word." That one tells the truth, sets expectations, and offers the door. Same disclosure. Completely different feeling.
The tone of the disclosure is part of the disclosure. Warm and plain says "we've got nothing to hide." Stiff and legal says "our compliance department made us do this." You want the first one.
Where do the rules and the law fit in?
Treat the law as your floor, not your finish line. Different regions are moving at different speeds. The EU AI Act, for instance, leans towards telling people when they're interacting with an AI system or seeing AI-generated content. Other places are further behind, and the rules will keep shifting for years.
So don't build your whole approach around one jurisdiction's current wording, especially if you serve customers in Britain, Australia, and Singapore all at once. Build it around the honest principle, don't let people be misled about whether they're dealing with a person or a machine, and you'll usually clear the legal bar in every market without redrafting every time a regulator updates a clause.
If you operate somewhere with specific obligations, check them. But the business that's already being straight with customers rarely gets caught out by a new rule. The ones that get caught are the ones who were hiding the ball.
How do you roll this out without overthinking it?
Start small, write it down, and tell your team. Three steps, none of them dramatic.
First, map every place AI meets a customer, chat, phone, email, decisions. Second, draft the one-line disclosure for each spot, in plain language, with a route to a human. Third, brief whoever's on the front line so they're not blindsided when a customer asks "wait, am I talking to a robot?" The right answer is a relaxed "partly, want me to grab a colleague?", not a panic.
In my own businesses, the principle has always been the same whether it's property management at EzyTrac or the counter at Darra Tyres: be the kind of business people trust because you tell them the truth before they have to ask. AI doesn't change that rule. It just gives you more places to honour it.
If you'd like a clear-eyed look at where AI touches your customers and a simple disclosure policy that fits how you actually work, book a free AI audit with Anaboo. No jargon, no hard sell, just an honest map of where you stand and what's worth doing next.
Live with passion & AI,
Brett
Need an AI operator inside your team?
Place a Chief AI Officer, an AI Officer, or embed an Anaboo Forward Deployed Engineer for 3–6 months.
Frequently asked questions
Do I legally have to tell customers they're talking to AI?
+
It depends on your region and use case, but the safe answer is to disclose anything that could mislead a customer about whether they're dealing with a person or a machine, and to check local rules like the EU AI Act if you trade there.
Will disclosing AI use scare customers away?
+
In practice the opposite tends to happen; a short, plain line telling people an assistant is helping out usually builds trust, while a customer who feels tricked into thinking AI was human is the one who walks.
Do I need to disclose AI used purely in the background?
+
No, you don't need to flag AI that drafts an internal report or sorts your inbox; disclosure matters when a customer is interacting with AI directly or when AI shapes a decision that affects them, like pricing or eligibility.
What's the simplest disclosure wording for a small business?
+
Something honest and human works best, such as "You're chatting with our AI assistant. Ask for a person any time and we'll hand you over", short, clear, and offering an exit.
How do I disclose AI without sounding like a legal disclaimer?
+
Write it the way you'd say it out loud to a customer, keep it to one sentence, put it where they'll actually see it, and always give them a fast route to a real person.
Brett is a four-time founder (Darra Tyres, Gladfish, EzyTrac, Anaboo) and the operator behind AIOS, Anaboo's AI Operating System. He writes from inside the build, installing AI in his own businesses first and reporting back what actually moves the numbers. Based between Singapore, the UK and Australia.
