AI transparency: building trust when your marketing uses AI

AI transparency means telling customers, simply and in the right place, when they are talking to a bot, when an image, voice or video was generated and presents something as real, and when their data is used to tailor what they see. It does not mean labelling every sentence an AI tool helped you polish. The rule of thumb is easy: if a customer would feel misled discovering it on their own, tell them first.

Why does AI transparency matter for a small business?

Because trust is a small business’s biggest asset. People buy from you because you feel reliable, reachable and real. The moment they learn that something presented as real was not, they start wondering what else was not.

The good news is that transparency costs almost nothing: one line in a chatbot greeting, a small label on an image, a sentence in your privacy notice. Hiding things, on the other hand, can cost a lot: an angry review, a screenshot doing the rounds on social media, or a regular customer who quietly stops coming.

There is a regulatory side too. Several markets already have, or are developing, rules that require businesses to tell people when they are interacting with an AI system or when content has been artificially generated. A business that builds the habit now will not have to scramble later.

When should you tell customers you use AI?

Whenever AI deals with the customer directly, or shows them something as if it were real. These are the four main cases.

When customers talk to a bot

This is the clearest case. Someone writing to you on WhatsApp or in your website chat should know whether a person or a digital assistant is answering. One line in the opening message is enough. Just as important, give them an easy way to reach a person. We explore that balance in our article on AI in customer service without losing the human touch.

When images or video present something as real

A dreamy illustration that is obviously imaginary misleads nobody. But an image that looks like a photo of your product, your hotel room or the result of a treatment must reflect what the customer will get. If AI generated it, say so.

When a voice or face is synthetic

A human-looking avatar, a synthetic voice in a video or on your phone line, a virtual presenter in a tutorial: all of these need a clear label. And never use a real person’s face or voice without their consent.

When customer data shapes what they see

If you use purchase history or past conversations to tailor messages and recommendations, explain it in your privacy notice in plain words, and let people opt out.

When is no disclosure needed?

When AI is a working tool and a person is responsible for the result. Drafting a post, fixing typos, translating text that someone checks, brainstorming headlines or summarising an internal meeting are much like using a word processor or a spellchecker. What matters is that someone in your business read it, corrected it and stands behind every word.

An illustration that is clearly an illustration, such as a cartoon character, needs no explanation either. Still, if your brand mascot doubles as your chatbot, it should introduce itself as a digital assistant whenever it talks to customers.

How do you disclose AI without putting people off?

Keep it short, friendly and useful. Transparency should not sound like a legal warning. It should sound like an introduction. A few examples:

A garden centre’s WhatsApp bot: “Hi, I’m the garden centre’s digital assistant. I can check stock, share opening hours and book a delivery slot. Want a person? Just type ‘team’ at any time.”

An image on a product page: “Visualisation created with AI. All other photos on this page are real.”

A privacy notice: “We use your order history to recommend products that suit you. You can ask us to stop at any time.”

A phone assistant at a clinic: “Hello, you’ve reached the clinic’s automated assistant. I can book or move an appointment. To speak to reception, press zero.”

Notice that in each example, the line about AI arrives together with something useful: what the assistant can do, and how to reach a human.

What happens when you don’t disclose?

Picture a boutique hotel whose WhatsApp bot signs off as “Emma from reception”. A guest chats with Emma before the trip, gets restaurant tips and feels a personal connection. At check-in she asks to meet Emma, and there is no Emma. Technically the bot did its job well. To the guest, she was fooled, and now every tip she received looks suspect too.

The very same bot, introduced differently, would create the opposite experience: “Hi, I’m the hotel’s digital concierge.” The guest gets the same tips at the same speed and walks away feeling well looked after, not tricked. The difference is a single sentence.

What you should never do

Some things cannot be done transparently, because all their value comes from deception:

• Reviews or testimonials written by AI and presented as genuine customer feedback
• A bot that uses a person’s name and insists it is human
• Before and after images that were generated or enhanced with AI
• Using a real person’s voice or face without explicit consent
• Hiding the option to talk to a person

How to build a simple AI transparency policy

You do not need a twenty page document. A short process the whole team knows is enough:

1. Map where AI meets customers. Chatbot, content, images, emails, recommendations, phone answering.
2. Decide for each point. Disclose, no disclosure needed, or do not use AI there at all.
3. Write standard wording. A chatbot greeting, an image label, a line for the privacy notice.
4. Add a short page to your website. How you use AI and how it serves your customers.
5. Brief the team. What is allowed, what is not, and who to ask when unsure. Our guide to an AI policy for a small business can help.
6. Review every few months. New tools creep into everyday use, and they need to be covered too.

What does AI transparency look like in practice?

A dental practice runs a WhatsApp bot that introduces itself, books appointments and sends pre-treatment instructions. For any clinical question it says plainly that this is one for the dentist and passes the conversation to the team.

An online jewellery shop labels its AI generated lifestyle shots, such as a necklace on a beach at sunset. The product photos themselves are shot in a studio, so buyers know exactly what they will receive.

An estate agency shows empty flats with virtually staged furniture, labels every staged image, and places a photo of the room as it really is right next to it.

An accounting firm records and summarises client calls with AI. Clients are told in advance and can ask for a call not to be recorded.

Quick AI transparency checklist

• Your chatbot introduces itself as a digital assistant in its first message
• A person is reachable at every step
• Realistic AI generated images are labelled
• Product photos show the real product
• Every review and testimonial on your site comes from a real customer
• Your privacy notice explains how data is used to tailor messages
• Your team knows what is allowed and who to ask

In short

AI transparency is not a confession. It is part of good service: customers know who they are talking to, know that what they see is real, and know there is always a person on the other side. At Kaede Digital, our own character Kaede introduces herself as an AI agent from the very first message, and we build our clients’ bots and assistants the same way: a clear introduction, an easy handoff to a person and well defined limits. You can read more about our AI agents for small business.

FAQ

Do customers trust a business less if it uses AI?

Usually not, as long as the use is open and helpful. People get annoyed when they feel misled or cannot reach a human, not because technology is involved. A digital assistant that answers quickly and hands over to a person when needed is generally seen as good service.

Do I have to label every post written with AI help?

No. When AI helps with drafting and a person in your business reviews, edits and takes responsibility for the content, it is much like using any other writing tool. Label content that presents something as real when it is not, such as a realistic AI generated image.

What should a chatbot’s first message say?

Three things: that it is a digital assistant, what it can help with, and how to reach a person. One or two sentences are enough. The shorter and friendlier it is, the more it feels like an introduction and the less like a disclaimer.

What if a customer is upset that they were talking to a bot?

Hand them to a person straight away, apologise simply and solve their problem. Afterwards, look at why it happened: perhaps the bot did not introduce itself clearly, or the route to a human was too hard to find.