Sooner or later, every business that uses AI in its marketing faces the same question: do we tell people? Some worry that admitting it will make their work look cheap. Others worry that staying quiet will look deceptive if it comes out later. Audiences, platforms and regulators are all paying more attention to the issue, and the expectations are still settling. What is clear is that trust is easier to keep than to rebuild.
This guide helps you decide when to disclose AI use, what to say, where to say it and how to build a simple policy your team can follow. It covers blog posts, images, emails, chatbots, reviews and customer communications. It is general guidance rather than legal advice, so check the rules that apply to your country, industry and platforms before you finalise your approach.
Key takeaways
- Disclose AI use when a reasonable reader would care, would be misled without knowing, or when a rule or platform policy requires it.
- Plain, short and specific wording works better than long legal language or vague claims.
- Disclosure does not replace quality. A person should still be accountable for the facts, the tone and the final decision.
- Never use AI to pose as a real person, invent testimonials or fabricate reviews, whatever the disclosure says.
- Write a one-page policy and apply it consistently, so decisions do not depend on who is publishing that day.
Why disclosure matters
Disclosure is about respect and trust. People want to know who or what they are dealing with, especially when decisions, money or personal information are involved. Being open also protects you. If a customer later discovers that content was machine-made and not mentioned, the surprise does more harm than the AI use ever would have.
Reasons to be open
- Trust: openness signals confidence and honesty.
- Compliance: some platforms, advertising codes and laws expect disclosure in certain situations.
- Accountability: saying that a named person reviewed the work shows who stands behind it.
- Expectations: audiences are better prepared for AI when they are told how it is used.
What audiences tend to care about
Attitudes differ between audiences and topics. People are usually relaxed about AI helping with routine tasks, and far less relaxed when it touches advice on health, money or safety, when it imitates a real voice or face, or when it is used to persuade them without their knowledge. Think about the stakes for the reader, not only about your own process. The higher the stakes, the more important it is to be open and the more important it is to keep a person involved.
Reasons businesses hesitate
Common worries include looking less authentic, giving competitors an idea of your process, or confusing readers. These worries can be managed with good wording. Most readers react well to a short, matter-of-fact note and badly to a feeling of being misled.
What counts as AI use worth mentioning
Not every use of AI needs a label. Spell checkers, grammar tools and translation of a single word do not usually require a notice. The question is how much AI shaped the content and whether the reader would care.
| Level of AI involvement | Example | Disclosure usually advisable? |
|---|---|---|
| Light assistance | Grammar or spelling corrections | Rarely |
| Editing support | Rewording your own draft for clarity | Optional, often a general statement |
| Substantial drafting | AI writes a first draft that a person edits | Often yes, in a general note |
| Generated media | Synthetic images, voices or video that look real | Yes, clearly |
| Automated interaction | A chatbot answering customers | Yes, at the start |
| Impersonation | Synthetic person, fake testimonial or review | Do not do it |
A useful rule of thumb is to disclose by function rather than by tool. Readers rarely care which product you used, but they do care about what it did. Saying that a draft was produced with AI and edited by a named person tells them what they need to know, while a list of software names does not.
The more a piece could be mistaken for the work or voice of a real person, the stronger the case for disclosure. Ask yourself the simple test: would the audience feel misled if they found out later?
When disclosure may be required
Rules vary and are developing. Platforms may ask you to label realistic synthetic media. Advertising bodies may expect clear identification of promotional content and honest claims. Some laws require that people be told when they are speaking to an automated system, or restrict deceptive synthetic content. Industry rules, such as those in finance or health, may add duties.
What to do about it
- List the platforms where you publish and read their current policies on AI content.
- Check the advertising and consumer protection rules in your markets.
- Note any contract terms with clients or partners about AI use.
- Review rules specific to your industry.
- Take advice from a qualified professional on anything high stakes.
This is general information, not legal advice. Because the rules change, review them regularly and keep a dated record of what you checked.
A simple decision guide
Use these questions in order. They are meant to be quick, so you can apply them to each piece of content.
- Is the reader interacting with a machine? If yes, say so up front.
- Is the content realistic media that could pass as real? If yes, label it clearly.
- Does a platform or rule require a label? If yes, follow it.
- Could the reader be misled about who wrote or created it? If yes, disclose.
- Does AI play a major role in the substance? If yes, a general statement is wise.
- Is AI only a light tool? Disclosure is optional.
Apply the guide at the point of publishing, not afterwards. If the answer is not obvious, ask a colleague who did not create the content whether they would feel misled. A fresh view is often the fastest way to spot a problem, and the conversation helps your team build a shared sense of what is acceptable.
When in doubt, lean toward saying a little more. A short note costs almost nothing, while an avoidable loss of trust costs a great deal.
Wording that works
Good disclosure is plain, specific and brief. Avoid alarming language and avoid apologising. State what happened and who is responsible.
Qualities of a good statement
- It says what AI did, such as drafted, translated or generated an image.
- It says what a person did, such as reviewed, fact-checked or edited.
- It uses everyday words rather than technical terms.
- It is short enough to be read.
- It appears where the reader will see it.
Patterns you can adapt
- Drafting: An AI tool helped draft this article. An editor checked the facts and revised the text.
- Images: This image was created with an AI tool.
- Chatbot: You are chatting with an automated assistant. You can ask for a person at any time.
- Translation: This page was translated with AI assistance and reviewed by a fluent speaker.
Tone matters as much as content. A disclosure that sounds embarrassed invites doubt, while one that sounds defensive invites argument. Write it the way you would explain any other part of your process, calmly and without drama. Test the wording on someone outside the business and ask what they understood from it.
Avoid claims you cannot support, such as saying content is fully human written when tools helped. Accuracy matters in both directions.
Where to place disclosures
A disclosure that nobody sees does not help. Put it where the reader makes the decision or forms the impression.
| Content type | Where to put the note | Style |
|---|---|---|
| Blog post | Near the byline or at the end | One or two sentences |
| Social post | In the caption or using the platform’s label tool | Short and visible |
| In the footer, or in the body if AI wrote the main message | Plain sentence | |
| Chatbot | First message and the chat header | Clear and immediate |
| Image or video | On or beside the media, plus platform label if offered | Visible, not hidden in fine print |
| Product pages | Near the AI-generated text or images | Specific |
| Customer support replies | At the start of an automated interaction | Offer a human option |
Accessibility matters as well. A label that appears only as a small icon or inside an image may not be read by people who use screen readers, so put important notes in real text. Keep the contrast high and avoid placing notes where they are likely to be cropped on mobile screens or by social platforms.
Consider a short page on your website that explains your approach to AI. You can link to it from your footer and from individual disclosures. It lets you say more without cluttering each piece of content.
Chatbots, automation and talking to people
Anyone talking to an automated assistant should know it. This is the clearest case for disclosure, because people share personal information and make decisions based on the conversation.
Good practice
- Say at the start that the assistant is automated.
- Offer an easy route to a human, and keep your promise.
- Do not give the assistant a fake human name and photo that suggest a real person.
- Be clear about what it can and cannot help with.
- Tell users how to find your privacy information.
For more on planning a support assistant that behaves well, see how to plan an AI customer support chatbot. The same honesty principles apply to any automated message sent on your behalf.
Reviews, testimonials and endorsements
This is a firm line, not a grey area. Do not use AI to invent customer reviews, fabricate testimonials, create fake case studies or write endorsements that a real person never gave. Consumer protection rules in many places treat fake reviews as deceptive, and audiences feel betrayed when they find them.
Safe and unsafe uses
| Use | Safe? | Notes |
|---|---|---|
| Tidying wording of a testimonial the customer approved | Yes, with approval | Keep the meaning unchanged |
| Summarising genuine feedback themes | Yes | Label it as a summary |
| Inventing a customer quote | No | Deceptive |
| Creating a fake persona as a customer | No | Deceptive |
| Drafting a request for reviews | Yes | Do not pressure or reward only positive ones |
The same care applies to stories about customers and to expert quotes in articles. If you use AI to draft a case study, base it on real information that the customer has confirmed and approved, and never invent details to make it more appealing. A real but modest story is far stronger than an invented impressive one, and it is safe if anyone ever checks.
If a person is quoted, the quote should come from that person. Keep records of consent and approval.
Writing a one-page disclosure policy
A short written policy keeps decisions consistent. It should sit alongside your wider AI use policy for your team.
What to include
- The principle: we are honest about how our content is made.
- The triggers: when a disclosure is required or expected.
- Approved wording for each content type.
- Who decides in borderline cases.
- What is never allowed, such as fake reviews or impersonation.
- Where records of review and approval are kept.
- How often the policy is reviewed.
Test the policy on three recent pieces of content before you adopt it. If your team disagrees about the answer for any of them, the policy needs clearer wording or an extra example. Keep a short list of worked examples at the end, because people learn from cases more easily than from principles.
Share the policy with everyone who publishes, including freelancers and agencies. Make it part of onboarding so nobody has to guess.
Keeping humans accountable
Disclosure is not a shield. Saying that AI helped does not excuse errors, bias or misleading claims. The person or business that publishes the content is responsible for it.
Minimum standards
- A named person reviews every public piece.
- Facts, figures and claims are checked against real sources.
- Tone and brand voice are reviewed.
- Sensitive topics receive extra care.
- Corrections are published openly when mistakes are found.
Our guide on how to check AI output for mistakes gives a practical review routine, and the discussion of writing blog posts with AI shows how to keep a real voice in the work.
Disclosure checklist
Keep this list close to your publishing workflow.
- Could a reader be misled about who or what made this?
- Is a person talking to an automated system, and have we said so?
- Is any realistic media labelled?
- Have we checked platform and advertising rules?
- Is the wording plain, specific and brief?
- Is the note placed where the reader will see it?
- Has a named person reviewed and approved the content?
- Are all testimonials and reviews real and approved?
- Have we recorded the decision?
Frequently asked questions
Do I have to disclose every use of AI?
Not necessarily. Light uses such as spelling and grammar checks rarely need a label. Disclose when AI substantially shaped the content, when realistic synthetic media is involved, when people interact with automated systems or when a rule requires it.
Will disclosing AI use hurt my credibility?
Usually less than being found out later. A short, confident note that explains how a person reviewed the work tends to be received well, and it shows that you take quality seriously.
Can I say content is human written if I used AI for ideas?
Be careful. Claims should be accurate. If AI contributed ideas or structure, saying nothing may be fine, but saying that no AI was used when it was would be misleading.
Do AI detectors prove whether content was machine written?
No. Such tools can be wrong in both directions, so do not rely on them to judge people or content. Focus on honest practice and your own records.
What about AI-generated images in ads?
If an image looks like a real photograph of a real person, product result or event, label it clearly and check advertising rules. Never use synthetic images to suggest results or endorsements that do not exist.
Who should decide our disclosure approach?
Assign one accountable owner, such as a marketing lead, and involve legal or compliance advice where needed. Review the approach regularly as rules and audience expectations change. Also check the terms of each AI tool you use, since they may carry their own rules about labelling.
Conclusion
Disclosing AI use is a habit, not a hurdle. Decide when a reader would care, use plain wording, place the note where it will be seen and keep a person accountable for the final result. Never use AI to fake people, reviews or endorsements. A simple policy applied consistently will take you further than any clever phrasing.
Rules and expectations will keep changing, so review your approach every few months and stay open with your audience. To build the policy into your wider practice, read how to write an AI use policy for your team, and keep quality high with how to check AI output for mistakes. Honesty and good editing are the best signals of trust you can send.
