AI Disclosure Statements for Businesses: What to Say and When

A useful AI disclosure statement tells customers what role AI plays in your work and where human responsibility sits. It should describe what you actually do, not repeat a broad promise that your business cannot support.

A general website statement, a note on an AI-assisted deliverable and a notice in a chatbot serve different purposes. This guide helps you choose the right one and offers wording to adapt.

These examples are practical guidance, not legal advice or a universal statement of disclosure law. Requirements depend on your location, industry, contracts and use of AI.

Start with the customer experience

Before writing a statement, ask:

1. Is the customer interacting directly with AI rather than a person?
2. Did AI substantially contribute to the work they receive?
3. Did AI meaningfully influence an important decision about them?
4. Would the customer reasonably expect to know about that use?
5. Do applicable rules or commitments require a particular explanation?

The answers help you choose the timing and detail. A background spelling suggestion is not the same situation as an AI agent answering a customer or an AI-informed recommendation affecting a person.

Canadian privacy regulators emphasize openness and transparency in their generative AI principles. Those principles are an important reference where personal information is involved, but they do not create a blanket rule that every business must disclose every use of AI. 1

Example 1: a general website statement

A public statement can explain your normal approach without listing every task or software feature.

We use AI tools to assist with drafting, research and summarization. Our team reviews material AI-assisted work before sharing it with customers. People remain responsible for our final work and decisions. If AI makes a substantial contribution where you would reasonably expect to know, we explain that use.

Only name activities your business actually performs. Do not claim that every output is reviewed if your process only covers some work. Explain confidentiality practices separately and accurately rather than making unsupported promises such as “no data ever leaves our systems.”

For an example of an organization publishing its approach, Destination Ontario has an AI disclosure statement covering its purpose, use cases, disclosure and responsibilities. Its institutional commitments are its own; they are not requirements that automatically apply to your business. 2

Example 2: a note on a customer deliverable

A website statement may not provide enough context for a particular document or recommendation. Where substantial AI assistance would matter to a reasonable customer, explain the role near the work itself.

AI assisted with the initial draft and summary in this report. Our team checked the material against the underlying sources, revised the content and approved the final report.

This is an example, not a claim about your workflow. If you did not check the sources, do not say you did. If AI also contributed analysis, describe that rather than calling it only drafting assistance.

Avoid vague labels such as “AI powered” when they obscure the actual contribution. A short explanation of the task and review is usually more informative than a product slogan.

Example 3: a customer-facing chatbot

People should not have to guess whether they are speaking to a person or an AI system. A notice at the start of the interaction can be simple:

You are chatting with an AI assistant, not a person. It can help with general product questions. To contact our team, use our contact form.

Adapt the contact method and limitations to what exists. Do not offer a human handoff if no handoff is available. Put the notice where the interaction starts rather than relying solely on a policy buried in a footer.

Example 4: an AI-informed decision about a person

Where AI meaningfully contributes to a consequential decision, a generic website disclosure may not explain that person’s situation.

An illustrative notice might say:

An AI tool helped organize information considered in this decision. A person reviewed the information and made the final decision. You can contact our team with questions about the process.

Use this only if it accurately describes the process, and check any additional legal or sector-specific obligations. Under EAUS, AI must not make final consequential decisions about people on its own.

Connect your statement to actual practices

The Ethicality AI Use Standard (EAUS 1.0) addresses meaningful AI use, substantial contributions to customer-facing work, direct AI interaction, consequential decisions and accurate public statements through requirements R5–R9. Meeting those requirements is not certification of legal compliance.

Pair disclosure with clear internal rules and meaningful human review. You can adapt the free transparency and disclosure templates, then take the readiness check.

The goal is not to attach an AI warning to everything. It is to give people accurate information when AI’s role matters to their understanding of the work or interaction.