Should You Tell Customers You Used AI?

The simple test

Ask one question: would the customer feel misled if they learned later that AI was involved? If yes, say so in plain language at the point where it matters, such as beside the chat box, under the image or in the decision notice. A notice buried in your terms does not count.

Customers are not against AI; they are against being tricked. A short, honest notice protects trust and gives your team a standard to apply before anything goes live.

When disclosure is expected or required

Chatbots and AI phone agents: people should know when they are not talking to a person. California's bot disclosure law already restricts undisclosed bots used to sell or to influence a vote, and Colorado's Chatbot Safety Act (HB 26-1263) requires operators of public conversational AI services to disclose that users are interacting with AI.
AI images of people or products: an image that is not a real photo of the actual product, person, place or result should say so.
Reviews and testimonials: never let AI invent reviews, quotes, ratings or customer stories. The FTC's rule on consumer reviews and testimonials, in force since October 2024, covers fake reviews, including AI-generated ones.
Decisions about people: hiring, lending, housing, insurance, pricing or eligibility decisions influenced by AI carry notice duties in some states and fairness duties everywhere.
Regulated industries: health care, finance, insurance, legal, education and housing already have advertising, consent and record rules that cover AI output.

Laws are changing quickly and vary by state. Check current official sources and get legal advice for your situation; this guide is not legal advice.

Colorado's new AI laws

In 2026 Colorado repealed and replaced its 2024 AI Act. The Automated Decision-Making Technology Act (SB 26-189) covers technology that materially influences consequential decisions about consumers, with notice and human-review rights. The Chatbot Safety Act (HB 26-1263) sets duties for public conversational AI, including disclosing that it is AI.

Both are scheduled to take effect on January 1, 2027, and the Colorado Attorney General is writing the rules. Its AI rulemaking page is the place to check current requirements. Businesses serving Colorado customers should inventory where AI touches decisions about people now.

When disclosure is optional

Spell check, grammar fixes, formatting and light editing that does not change meaning usually need no label. A bakery does not need to flag that a tool fixed the commas in its menu.

Optional is not careless. If AI rewrites a message so much that it adds claims, changes a promise or makes expertise sound broader than it is, review it and consider a notice. The line is not whether a tool was used; it is whether AI changed what the customer understands. Watch for invented details; our AI hallucinations guide explains how to catch them.

Sample disclosure wording

Website chat: "Hi, I'm an AI assistant. I can answer common questions and route your request. Type 'person' at any time to reach our team." Higher-risk businesses can add: "I can't give medical, legal or financial advice."
Phone agent opening: "You're speaking with an automated assistant for [business]. Say 'representative' to reach a person."
Email footer: "This message was drafted with help from AI and reviewed by our team." Use it only if a person really reviewed it.
Product image: "AI-generated image for illustration. See the gallery for photos of the actual product."
Decision notice: "An automated tool helped review this application. You can ask a person to review the decision by contacting [contact]."

Reviews and testimonials: the hard line

AI may fix formatting or spelling in a real customer's words, with permission, but the meaning must stay theirs. Adding enthusiasm, new claims or removing their caveats turns editing into fabrication.

Keep the original statement on file.
Get permission for any edit.
Never add performance claims the customer did not make.
Disclose relationships such as payment, free products or employment.
Do not use AI avatars to imply a real person or celebrity endorsed you.

Build disclosure into your workflow

Inventory customer-facing AI: chatbots, phone agents, image tools, review summaries, recommendation engines, lead scoring and decision tools.
Rate the risk: spell check is low; automated eligibility decisions are high.
Pick the disclosure point: chat window, call opening, image caption, testimonial block, email footer or decision notice.
Name an owner: someone approves AI-generated claims, images and decision flows before customers see them.
Keep records: sources, approvals and permissions for higher-risk uses.
Review quarterly: laws, platforms and tools change.

Put the rule in your AI acceptable use policy so staff apply it the same way.

Where ASL fits

A disclosed assistant is only as good as its answers. Viavitna, our AI knowledge base for customer questions, answers only from content you approve, cites its sources and declines when your content does not cover a question. Read who pays when a chatbot makes a promise, or contact us to review your customer-facing AI.

Frequently asked questions

Do I have to tell customers I use AI?

Not for spell check or light editing. You should disclose when customers talk to a bot, see AI-generated people or products, read AI-assisted reviews or receive decisions influenced by AI, and some laws require it.

Is it legal to use AI-written reviews?

Fake reviews are not. The FTC's rule on consumer reviews and testimonials covers reviews by people who do not exist, including AI-generated ones, and reviews by people with no real experience of the product.

What does Colorado require?

Colorado's Chatbot Safety Act and Automated Decision-Making Technology Act, scheduled for January 1, 2027, add AI disclosure and notice duties. Check the Colorado Attorney General's AI page for the current rules.

What should a chatbot disclosure say?

That the customer is talking to an AI assistant, what it can help with and how to reach a person, placed at the start of the conversation.

Will disclosing AI scare customers away?

Clear, short disclosure usually builds trust. Customers react badly to finding out later that a person, photo or review was not real.

Related guides

Explore the complete product

Read features, current access and limitations