What Is an AI Disclosure and How a Business Can Do It Right
AI disclosure tells customers where and how your business uses AI. What the law asks, what to disclose, how to write and place it, and how to make it checkable.

AI disclosure is the practice of telling customers where and how a business uses artificial intelligence in the products and services they interact with. The practice has moved from optional to expected, because customers increasingly assume a business uses AI somewhere and want to know where. In some places and for some uses, it is now also a legal requirement.
This guide explains what AI disclosure covers, where the law already asks for it, what a business needs to disclose, how to write a disclosure statement, how to present it to customers, and the step that separates a disclosure customers believe from one they quietly doubt. The aim is a clear account a business can act on, written for the people running that business rather than for a compliance file.
What AI disclosure is
AI disclosure is a written statement in which a business identifies the ways it uses artificial intelligence in its dealings with customers. The statement records each use of AI that bears on a customer, describes what the system does, and sets out what the customer may expect as a result. Where a privacy policy governs the handling of personal data, an AI disclosure governs the conduct of the automated systems that act on that data.
A privacy policy and an AI disclosure answer different questions, and businesses often treat one as if it covers the other. Most standard privacy policies were written before AI shaped customer experiences this directly, which is why they leave the AI questions unanswered. A complete approach uses both documents.
Privacy policy | AI disclosure | |
|---|---|---|
What it covers | The data a business collects, stores, and shares | The automated decisions that data is used to make |
Question it answers | What do you know about me? | What does AI decide about me? |
Worked example | We collect your purchase history | Your purchase history feeds a system that sets the products and prices you see |
Governs | The handling of personal data | The conduct of the automated systems acting on that data |
On a website, AI disclosure also sits next to tracking and consent. A chat widget, a personalization tool, or an AI-driven ad script usually loads third-party code in the visitor's browser, and some of that code can set cookies or send data before the visitor has made a choice. The same review that finds your AI uses should check what those tools load and whether your consent banner controls them.
Where the law already asks for AI disclosure
There is no single AI disclosure law. A handful of rules, each with its own scope, already require some form of it. The summaries below describe what each rule asks and who it applies to. They are not legal advice; whether a rule covers your business depends on your facts.
United States, FTC Act. The Federal Trade Commission applies its existing law on unfair and deceptive practices to AI. In September 2024 it announced Operation AI Comply, a set of actions against companies that made deceptive AI claims, with then-Chair Lina Khan stating that "there is no AI exemption from the laws on the books." The FTC's rule on consumer reviews and testimonials (16 CFR Part 465, announced August 2024) bans fake reviews, including reviews that misrepresent that they come from someone who does not exist, such as AI-generated fake reviews. In practice: what you say about your AI, and what your AI says on your behalf, must be true.
California bot disclosure law (Bus. & Prof. Code 17940-17943). It is unlawful to use a bot to communicate or interact with a person in California online with the intent to mislead them about its artificial identity in order to encourage a purchase or sale, or to influence a vote. A person who discloses that the communication comes from a bot is not liable, and the disclosure must be clear, conspicuous, and reasonably designed to inform the people the bot interacts with (section 17941). "Online" covers any public-facing website or app (section 17940).
California AI Transparency Act (SB 942). This law is aimed at companies that create or produce generative AI systems publicly accessible in California, not at businesses that only use them. It requires those providers to offer a tool for checking whether content came from their system and to embed provenance information in the content. It was amended in 2025 (AB 853) and again in September 2026 (SB 1000), which removed the original one-million-user threshold (Morrison Foerster summary, October 2026).
Colorado (SB 26-189). Colorado's 2024 AI Act (SB 24-205) was repealed and replaced by SB 26-189, signed on May 14, 2026 and scheduled to take effect on January 1, 2027 (Mayer Brown summary); a federal court challenge brought by xAI is pending. It covers "covered automated decision-making technology" that materially influences consequential decisions in education, employment, housing, lending, insurance, health care, and essential government services. Businesses using it must give consumers clear and conspicuous notice at the point of interaction, and after an adverse outcome, a plain-language description of the technology's role within 30 days.
Utah AI Policy Act. Since amendments that took effect on May 7, 2025, a business using generative AI with a consumer must clearly and conspicuously disclose that the consumer is interacting with AI if the consumer clearly and unambiguously asks. People in state-licensed occupations must disclose at the outset of "high-risk" interactions, such as those that collect sensitive personal information and give personalized financial, legal, medical, or mental health advice. The act is set to run until July 2027 (Davis Polk summary).
European Union, AI Act Article 50. The transparency obligations of Regulation (EU) 2024/1689 apply from August 2, 2026. Providers must design AI systems that interact directly with people so those people are informed they are interacting with AI, unless that is obvious. Deployers must disclose deepfake images, audio, and video, and AI-generated text published to inform the public on matters of public interest, unless the text has had human review and someone holds editorial responsibility. The information must be clear and distinguishable and given at the latest at the first interaction or exposure (Article 50 text).
The common thread: tell people when they are dealing with a machine, do not let AI mislead them, and explain automated decisions that affect them. A disclosure written to the standard below covers that ground in plain language.
What a business needs to disclose
A business discloses any AI use that affects a decision about the customer, their money, their data, or their access to a service. That standard draws a clear line. Uses that touch a customer directly belong on the disclosure. Internal tools that a customer never encounters do not.
The test for each AI use is one question. Does this system affect a decision the customer would want to know about? A pricing system that adjusts figures based on demand affects the customer, so it goes on the disclosure. A recommendation engine that determines which products a customer sees affects the customer, so it goes on the disclosure. An automated check at checkout that can hold or decline an order affects the customer, so it goes on the disclosure. A support system that ranks and routes requests affects the customer, so it goes on the disclosure as well.
The systems that stay off a public disclosure are the ones a customer never experiences. An AI tool that schedules warehouse staff makes no decision about a customer, so it does not require a public statement. The line holds wherever AI shows up in an operation. A system that decides something about a person belongs on a disclosure that person can read. A system that runs in the background of internal operations does not.
AI use | Affects a customer decision? | Disclose it? |
|---|---|---|
Pricing that shifts with demand | Yes | Yes |
Product recommendations | Yes | Yes |
Automated order screening at checkout | Yes | Yes |
Support chatbot | Yes | Yes |
Warehouse staff scheduling | No | No |
Internal sales forecasting | No | No |
How to write an AI disclosure statement
An AI disclosure statement is the page a customer reads to understand how a business uses AI. Written well, it answers the questions in a customer's head before they have to ask. Written in vague or legal language, it raises more doubt than it settles. The standard is plain language at a level any customer can read, with one short entry for each AI use the business has identified.
What to include in the statement
A disclosure statement documents each customer-facing use of AI as a separate entry. Every entry addresses three elements:
Function. The specific task the system performs. Name it precisely. A statement that a business uses AI "to improve the customer experience" discloses nothing, because it identifies no system and describes no behavior.
Purpose. The business reason the system is in use, which gives the customer the context to judge whether that reason is reasonable.
Effect on the customer. What the system does to the customer's experience, and any control the customer holds over it, such as the ability to opt out or request human review.
A well-drafted entry also anticipates the customer's primary concern about that particular system and addresses it directly.
For a dynamic pricing system, the concern is whether the customer is being charged more than others; the entry should state whether personal data influences the price shown. For an automated support channel, the concern is whether the customer knows they are not speaking to a person; the entry should state plainly that the assistant is automated. Naming the concern and answering it is what separates a disclosure that builds confidence from one that merely satisfies a formality.
The completeness of the statement is judged by a single standard. A customer should finish it able to identify every point at which AI affects them, with no material use omitted and no ambiguity about what each system does.
Where to place it on your website
A disclosure a customer cannot find provides little value. The placement should match the moment a customer would want the information.
A link in the main footer makes the full statement reachable from every page, which is the baseline.
A short line near a chatbot window noting that the assistant is automated, shown before or at the first message.
A one-sentence note on a pricing page about demand-based pricing, linked to the full statement.
Context-specific notices at checkout or in account emails, wherever another AI use occurs.
The full page carries the detail, and the in-context notices reach the customer at the exact moment the question would occur to them. To see whether your site already shows an AI disclosure and what loads before consent, run the free Trusted Site Scan.
How to present AI disclosure to customers
A disclosure is also something your team says out loud, in emails, sales calls, and support chats. How it is presented decides whether it reads as candor or as fine print.
Lead with what the AI does for the customer. "We use AI to suggest products based on your past orders. Here is exactly how" says more than "We use AI to optimize recommendations."
Keep the wording consistent across channels. The footer page, the chatbot notice, the checkout note, and what staff say should describe the same uses in the same terms.
Label AI-drafted messages. If AI drafts emails or outreach, a short line such as "This message was drafted with AI assistance" avoids the feeling of a fake personal note.
Show where people stay in charge. Say which steps a person reviews or decides, and how a customer reaches a human. Customers want to know someone is accountable for what they receive.
Give staff an answer ready. Sales and support teams should have a short, accurate answer for "Do you use AI for this?" and "How do I know that is true?" A confident, specific answer builds trust; an awkward deflection does the opposite.
Disclosure is not risk-free. Some customers will prefer a business that uses less automation, and a few may leave. The alternative is worse: customers who discover undisclosed AI on their own tend to ask what else was not said. You can read more about what customers want to hear in what customers actually want to know about your AI use.
Why disclosure alone is often not enough
A disclosure statement establishes what a business says about its AI use. It does not establish whether that statement is true. A customer reading a disclosure has only the company's word that the account is complete and accurate, and the company writing the description is the same company that benefits from appearing transparent. That structure limits how much a self-written disclosure can prove.
This is where writing your own claims, without outside review, shows its weakness. A business using AI carelessly can publish a disclosure that reads identically to one published by a business using AI well. From the outside, a customer cannot tell the two apart, because the words are all they have to judge. A polished statement and an honest practice are not the same thing, and a disclosure on its own cannot show which one a customer is looking at.
Unverified disclosure can also work against the business that publishes it. A statement that a company uses AI responsibly, offered with nothing to confirm it, invites a customer to ask what the company chose not to say. A vague statement can leave a customer more skeptical than no statement at all.
How verified disclosure works
Verified disclosure closes the part a self-written statement cannot. It adds two components to the disclosure itself: outside review and a public record.
Outside review means a party other than the business checks the disclosure against what is actually on the site and in its systems, and against a defined set of criteria. A statement someone else has examined carries more weight than a self-written claim.
A public record lets anyone check the result without depending on the company's own words. A customer or partner looks up the business and confirms its status directly.
A badge that links to a live verification page can be checked. A badge that exists only as an image asks to be trusted. SiteTrust works this way: it scans a site for AI disclosure, tracking before consent, and legal pages, publishes the business's AI policy, and gives it a trust badge that links to its entry in a public registry. For the difference between checked and self-declared claims, see AI transparency verification vs self-attestation.
Why verified disclosure can become an advantage
Verification gives the customers who research a business before buying something to check, rather than an assurance to accept. It separates a business from competitors who offer only self-claims, because a badge that can be checked stands apart from one that cannot. And it prepares a business for the rules above, which are moving toward more disclosure, not less.
The advantage builds over time. Each customer who checks a business and proceeds is a customer the business might have lost to doubt under a self-written claim. To see where your site stands today, start with the free Trusted Site Scan, or compare the plans on the pricing page.
Frequently asked questions about AI disclosure
Is an AI disclosure the same as an AI policy?
They overlap. An AI disclosure tells customers where AI affects them. An AI policy is usually broader: it also sets the rules a business follows when it uses AI, such as human review, data handling, and how errors are corrected. Many businesses publish one page that does both. The AI policy page shows what that can look like.
Does a small business with one chatbot need an AI disclosure?
A chatbot is the most common case where the law already expects one. California's bot law makes it unlawful to use a bot to mislead someone about its artificial identity to encourage a sale, and Utah requires disclosure when a consumer asks. The EU AI Act requires chatbots to be designed so people know they are talking to AI. A one-line notice in the chat window plus a link to a fuller statement is a small cost for that coverage.
Do I need to label AI-written blog posts or AI images?
It depends on the content and the market. In the EU, deployers must disclose deepfake images, audio, and video, and AI-generated text published to inform the public on matters of public interest unless a person reviewed it and holds editorial responsibility. In the US there is no general labeling rule for marketing content, but the FTC Act's ban on deceptive practices still applies: AI-generated reviews, testimonials, or images must not mislead customers about what is real.
How does AI disclosure relate to cookie consent?
They live side by side on the same pages. Many AI tools on a website, such as chat widgets and personalization scripts, are third-party code that can set cookies or send visitor data. Your AI disclosure explains what those tools do; your consent banner decides whether they load before the visitor agrees. Checking both together avoids a disclosure that describes a tool your banner does not control.
How often should an AI disclosure be updated?
Update it whenever you add, remove, or change a customer-facing AI tool, and review it on a fixed schedule, for example every quarter, even if nothing seems to have changed. Vendors add AI features to existing tools without much notice, so a review should check the tools you already have, not only the new ones.
What happens if a disclosure says something that is not true?
An inaccurate disclosure can be worse than none. In the US, a false or misleading statement about how a business uses AI can be treated as a deceptive practice under the FTC Act, and state consumer protection laws work the same way. Keep the statement specific, keep it current, and do not describe safeguards you do not actually have.
Ready to become a founding member?
Apply for certification todayFree AI tools
Assess your AI readiness
Use these quick assessments to spot trust, governance, and disclosure gaps.









