How to Write an AI Disclosure Statement for Your Website

Say plainly where you use AI (for example chat replies, product descriptions or room mock-ups), where you don't, that a person checks what customers see, how customer data is handled, and how to reach a human. Keep it to 150-250 words on its own page, linked from your footer, chat window and product pages.

The page is the reference; the short notices at the moment of use, such as the first line of a chat or a label under an AI-generated image, matter more. Some disclosures are legal duties, not courtesies, particularly if you sell to customers in the EU. Below you'll find an inventory of where customers meet AI, the rules that may apply, a fill-in template, a complete example for a picture framer, and the in-context wording to go with it. Whether to disclose at all is a separate question, covered in should you tell customers you use AI?; this tutorial is about writing the words once you've decided.

Follow me on Instagram@sagnikteaches

Start with an inventory of where customers meet AI

You can't write an accurate statement without an accurate list. Walk through your website and customer processes and note every place AI is involved. For the picture framer in the worked example below, it looked like this:

Connect on LinkedInSagnik Bhattacharya

Keep this list. It's the source for your statement, and when you add or drop a tool, it's the first thing to update.

Subscribe on YouTube@codingliquids

A service business's list can look quite different. For an illustrative two-person wedding-cake studio, the inventory had only three lines: AI-drafted replies to enquiries (checked before sending), Instagram captions (edited), and AI-generated design sketches made from a couple's mood board. The sketches were the line that mattered. They looked like photos of finished cakes, and one couple had assumed a sketch showed a cake the studio had already made. The fix was a label on every sketch ("AI design sketch, not a cake we've made") and one sentence in the statement, which is also the kind of realistic, could-pass-as-authentic image the EU rules below care about.

When disclosure is a legal duty rather than a courtesy

This isn't legal advice, and the rules depend on where your customers are. Three sets of rules are worth knowing about:

• The EU AI Act, if you sell to customers in the EU. Its transparency rules in Article 50 have applied since 2 August 2026. People must be told they're interacting with an AI system, from the first interaction, unless it's obvious; that duty falls mainly on the provider that builds the chat tool, but the notice appears on your site, so check it's there. Businesses using AI must disclose "deepfakes", meaning AI-generated or manipulated images, audio or video that resemble real people, places or objects and could pass as authentic. And AI-generated text published to inform the public on matters of public interest must be labelled, unless a person has reviewed it and someone holds editorial responsibility. The EU's official FAQ on Article 50 transparency obligations sets this out in plain terms.
• Google's rules for online shops. Google's guidance on generative AI content says AI-generated product data such as titles and descriptions must be specified separately and labelled as AI-generated in Merchant Center, and AI-generated images must carry the IPTC DigitalSourceType metadata value TrainedAlgorithmicMedia. If you list products on Google, check your product feed settings.
• Data-protection law. If an AI tool processes customer information, for example chat transcripts stored by a chat vendor, your privacy notice needs to describe it. The disclosure statement can link to it but doesn't replace it. What to check in an AI tool's privacy policy and terms tells you what to find out from the vendor first.

Beyond these, general consumer rules against misleading customers apply everywhere, which is why statements that overclaim are risky (see the wording table below). If you're in a regulated profession, check your regulator's guidance too, and ask a solicitor if you're unsure how any of this applies to you.

Five parts of a good statement

1. Where we use AI: specific uses, in customers' terms. "Our website assistant answers questions about sizes and prices," not "we use AI technologies".
2. Where we don't: the human parts customers care about. For a framer, that's the craft: measuring, cutting, mounting and fitting.
3. Who checks what: which AI outputs a person reviews before customers see them, and which (such as live chat replies) they don't.
4. Your data: what information the AI tools see, whether it's used to train AI models, and a link to your privacy notice.
5. How to reach a person: a phone number, email address or a way to switch from the chat to a human.

Add a "last updated" date. It tells customers the statement is maintained, and it reminds you to maintain it.

A template you can fill in

HOW WE USE AI
Last updated: [date]

We use AI tools to help with some parts of our work, so we can
[reply faster / keep prices down / spend more time on X].
Here's exactly where.

Where we use AI
- [Use 1, in plain words, e.g. "Our website assistant is an AI.
  It answers common questions about ..."]
- [Use 2, e.g. "Some product descriptions are drafted with AI
  and edited by our team."]
- [Use 3]

Where we don't
[The human work customers value, e.g. "Every frame is measured,
cut and fitted by hand by our framers."]

What people check
[e.g. "A member of our team reads every email and description
before it's sent or published. Our website assistant replies
automatically, so if anything it tells you matters to your
order, we'll confirm it in writing."]

Your information
[What the AI tools see; whether it's used to train AI models;
"See our privacy notice for details: [link]"]

Talking to a person
[Phone / email / how to switch from the assistant to a person.]

AI can draft the statement from your inventory, but check the draft for promises you never made. A prompt that works:

Here is a list of where our business uses AI and where customers
see its output: [PASTE INVENTORY]. Write a "How we use AI" page
of 150-250 words using these headings: Where we use AI, Where we
don't, What people check, Your information, Talking to a person.
Plain words, no marketing language. Only describe the uses in
the list. Where you don't know something (such as how long a
vendor keeps data), write [CHECK] instead of guessing.

An illustrative extract from what comes back:

"Our website assistant is an AI that answers questions about sizes and prices. All of our content is carefully reviewed by our expert team, and we comply fully with all applicable AI regulations. Your data is never shared with third parties."

The first sentence is fine. The other two are exactly the overclaims in the wording table below: live chat replies aren't reviewed, "comply fully" is a legal claim nobody has checked, and a chat vendor storing transcripts is by definition a third party. The [CHECK] instruction helps, but it doesn't stop the tool filling gaps with reassuring boilerplate, so read every sentence against the inventory.

Worked example: a picture framer's statement

To illustrate, imagine a framing business with a high-street shop, a workshop and an online shop, where AI is used in the ways listed in the inventory above. Its statement, at about 210 words:

How we use AI. Last updated: [month, year].

We use a few AI tools so we can answer questions quickly and spend more time at the workbench. Here's where.

Where we use AI. The assistant in our website chat is an AI. It can answer questions about frame sizes, prices and turnaround times, and it'll pass you to a framer if you ask. Our "see it on your wall" previews are generated by AI from your photo. Some of our product descriptions and email replies are drafted with AI, then edited and checked by our team.

Where we don't. Every frame is measured, cut, mounted and fitted by hand in our workshop. Our product photos show the real frames; we only use AI to tidy backgrounds.

What we check. A person reads every email and description before it goes out. Chat replies are automatic, so we confirm prices and dates in writing before we start work. Previews are approximate: colours and scale can differ, and we'll show you real samples before cutting.

Your information. Our chat provider stores conversations for [period] to run the service and doesn't use them to train AI models. Our privacy notice has the details.

Talking to a person. Call or email us, or type "framer" in the chat.

Before publishing a line like "doesn't use them to train AI models", confirm it in the vendor's current terms. A disclosure statement that's wrong about data is worse than none.

The short notices that matter more than the page

Few customers read a disclosure page. Most will see a one-line notice at the moment of use, so write those with the same care:

• Chat opening line: "Hi, I'm the [shop name] assistant, an AI. I can help with sizes, prices and turnaround. Type 'framer' at any time to reach a person." The tutorial on what to tell customers at the start of a chat has more variations.
• Under an AI preview: "AI-generated preview. Colours and scale are approximate; we'll confirm with real samples before cutting."
• Email footer (optional): "Some of our replies are drafted with AI and checked by our team."
• Phone line answered by AI: "You're through to the [business] booking assistant, an automated AI service. I can book or move appointments. Say 'person' to leave a message for the team."
• Social posts with AI images: follow the platform's labelling tools; how to label AI-generated images and video on social media covers them.

Once the notices are live, test them the way a customer would, because the assistant's own answers can contradict the page. Open the chat on your phone and ask three questions: "Am I talking to a real person?", "Do you keep what I type here?" and "Can you guarantee my frame will be ready by Friday?" An illustrative set of answers from a freshly set-up assistant:

"I'm part of the team here and happy to help! Your privacy is important to us and your details are kept completely secure. Yes, we can have your frame ready by Friday."

All three are problems. The first dodges the question the opening line was supposed to answer. The second is vague reassurance the statement never made. The third is a promise the assistant can't keep, since turnaround depends on the workshop's queue. The fix goes in the assistant's instructions, not on the disclosure page: "If asked whether you're a person, say you're an AI assistant. For questions about data, link to the How we use AI page. Never confirm a completion date; say a framer will confirm it in writing." Re-run the same three questions after every change to the instructions.

Wording that creates problems

How the last row goes wrong, in an illustrative hair salon: the chat assistant was given a friendly first name and a stock photo of a smiling stylist, with "AI" mentioned only on the disclosure page. A customer spent ten minutes arguing about a cancellation fee, then asked to speak to the manager "about what your receptionist told me". She was surprised, and not pleased, to learn she'd been talking to software. Keeping the name but adding "(AI assistant)" beside it, and dropping the photo, cost nothing.

The same before-and-after logic works on the whole statement. A before, from an illustrative heating engineer's footer:

"Powered by AI for faster, smarter service."

And after:

"Outside office hours our booking line is answered by an AI assistant. It can book, move or cancel a service visit, and it'll tell you if your problem needs a call back from an engineer. Everything it books is checked by our office the next morning."

The first says nothing a customer can rely on. The second tells them what the AI does, what it can't decide, and when a person looks at it.

Reviews and testimonials need particular care: AI can help you ask for and reply to reviews, but not write them. The tutorial on using AI to write testimonials explains why.

Where the statement should live

Give it a page of its own with a plain address, such as /how-we-use-ai, and a footer link labelled "How we use AI". Then link to it from the places where customers actually meet AI:

• the chat window's menu or first message;
• the preview tool, next to the "AI-generated preview" label;
• your privacy notice, in the section about service providers;
• your "about" or "how we work" page, if you have one.

Write it as ordinary text on the page, not as an image or a downloadable file, so screen readers and search engines can read it. Article 50 of the EU AI Act also expects the information to meet accessibility requirements, and plain text is the simplest way to do that.

Keeping the statement true

A disclosure statement goes out of date the moment you add a tool, and an out-of-date one is misleading. Tie it to your other processes:

• Whenever a new AI tool is approved, update the inventory and the statement in the same week.
• Review it alongside your internal AI policy every six months; writing an AI usage policy sets up that review cycle.
• Re-check vendor data terms at renewal, since the "your information" section depends on them.
• Update the "last updated" date each time, even if the change is small.

The quarterly check is where drift turns up. At the framer, the third review found two things. The chat vendor had added an option to "improve our AI with your conversations", switched on by default for new widgets, so the "doesn't use them to train AI models" sentence was no longer true until the owner switched it off. And a member of staff had started using an AI tool to recolour customers' photos for a new "frame your pet portrait" service, a use that wasn't on the inventory at all. One line went into the statement ("Pet portrait colours may be adjusted with AI; we'll send you a proof before printing") and the "last updated" date moved on. Neither change took long, but both would have left the page saying something false.

Done properly, the statement takes an hour to write and ten minutes to maintain each quarter, and it gives your team a straight answer when a customer asks, "Was that written by a person?"

Questions about AI disclosure pages

Do I need a disclosure statement if I only use AI behind the scenes?

Usually it's optional. If AI only helps with internal work, such as bookkeeping or summarising supplier price lists, and customers never see its output, there's little to disclose beyond what your privacy notice says about data. Many businesses still add a short statement, because customers increasingly ask and a clear answer builds trust.

Should I name the AI tools I use?

You don't have to, and naming them means updating the page whenever you switch. Describe the type of tool and what it does instead, for example an AI assistant that answers questions about sizes and prices. Name a vendor only where it helps customers, such as when their data is processed by that vendor and your privacy notice already names it.

Will saying I use AI hurt my search rankings?

There's no sign that disclosure itself harms rankings. Google's guidance says it doesn't object to AI-assisted content as such; what it penalises is content produced at scale with little value added. Its guidance also says that sharing how content was created can give readers useful context. Useful, accurate pages matter far more than whether AI helped write them.

Further reads

• Should You Tell Customers When You Use AI in Your Marketing? — Disclosure decisions specific to marketing content.
• How to Mention AI Use in Client Contracts and Proposals — Put the same promises into proposals and contracts.
• AI Mistakes That Damage Customer Trust, and How to Avoid Them — What undermines trust even with good disclosure.
• Who Is Liable When Your AI Chatbot Gets It Wrong? — Who is responsible when the assistant gets it wrong.
• AI Compliance Checklist for Small Businesses: What Applies to You — Other rules that may apply alongside disclosure.
• AI Product Photos vs a Photographer: Cost and Quality Compared — Weigh AI imagery against real photography first.
• AI Ethics for Small Businesses: A Practical Checklist — Twenty-five questions in six groups, each with why it matters and how to check, plus red lines and a nursery example run end to end.
• Should You Charge Clients Less When AI Speeds Up Your Work? — What to do about your prices when AI cuts the time a job takes: the rules for hourly billing, when fixed fees are fair, and how to tell clients.
• How Estate Agents Use AI to Write Property Descriptions — Build the fact sheet at the valuation, prompt in your house style, get portal, brochure and social copy from one input, and check every claim before listing.
• AI Virtual Staging for Property Listings: Costs and Disclosure — Per-photo and per-listing costs of AI virtual staging, the edits a label can't excuse, label wording to copy, and a pre-publish check for every image.
• Can AI Answer Calls and Qualify Leads for an Insurance Agency? — How an AI voice agent should route, qualify and score insurance calls, the phrases it must never say, a 20-call test plan and what it costs to run.
• Should Freelancers Tell Clients They Use AI? — Decide when to disclose AI use, ask for specific approval and explain your process without making promises you cannot support.
• Should Therapists Tell Clients They Use AI? — Which AI uses therapists must tell clients about, ready-to-adapt wording for contracts and first sessions, and what to do when a client says no.
• How Nurseries Use AI to Write Daily Reports for Parents — The drafting step from practitioner notes to parent's phone, built-in writing assistants, a house style card, and the wrong-child mistake to design out.
• Can You Use AI for Grant Writing Without Funders Rejecting It? — What funders have said about AI-assisted applications, why generic drafts score badly, and a worked $15,000 application that stays in your own voice.
• Charity AI Risks: Data Protection, Deepfakes, and Donor Trust — Three risk areas for charities using AI, with real cases, the controls that work for a small team, and a filled-in one-page risk register.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: the EU's official FAQ on transparency obligations under Article 50 of the AI Act; Google Search Central guidance on using generative AI content (including Merchant Center labelling for AI-generated product data and images). Checked September 2026. Not legal advice.