Honest Marketing: Claims, Disclosure, Copyright and AI Labels
AI makes marketing faster, and it makes honest mistakes faster too. A chat tool will happily write "clinically proven", "used by thousands" and "only three left" about a product that has no study, forty customers and plenty of stock. An image tool will show your product in a colour you do not make. A voice tool can imitate a person who never agreed. Each of these can mislead customers, break a rule or damage the trust you spend years building. This tutorial gives you a practical way to keep your marketing honest and lawful: how to test a claim, how to handle testimonials and paid endorsements, how to avoid false urgency and misleading prices, what to disclose and label, what copyright means for the images, music and words you use, how privacy rules apply to email and tracking and a checklist to run before anything goes live. It is general guidance, not legal advice, and it tells you where to check the rules that apply to you.
Difficulty: ★★☆☆☆ (no technical background needed)
Tools needed: Any chat AI, and the marketing material you are about to publish
Last updated: November 2026
Overview
Honesty in marketing is not only an ethical matter, though it is that. It is practical in three ways.
1. It is the law, in most places. Consumer-protection, advertising and competition rules exist in nearly every market, and they generally forbid misleading claims, fake reviews and hidden terms. Penalties and complaints are real, and small businesses are not exempt.
2. It protects trust. A customer who feels misled does not come back, leaves a review and tells friends. For a small business whose growth depends on reputation, that is expensive.
3. It protects your platforms. Advertising and social platforms reject or remove ads with unsupported claims, and can restrict accounts that repeatedly break their rules.
The central test is simple: would a reasonable customer be misled by this, and could I prove what I said? If the honest answer to the first is yes, or to the second is no, change it.
AI changes the risk in two ways. First, it produces fluent, confident marketing language by default, and that language is full of unsupported claims. Second, it makes creative that looks real: realistic product images, lifelike people and voices, convincing "reviews". Speed and polish make it easy to publish without checking.
So this tutorial builds a habit: check claims before they go live, keep evidence, label what needs labelling and use only what you have the right to use. It will also show you how to use AI itself as a checker, in a careful way.
The tutorial has seven steps. First, the claim test. Second, the main kinds of claim that cause trouble. Third, testimonials, reviews and endorsements. Fourth, urgency, prices and offers. Fifth, disclosure and AI labels. Sixth, copyright, likeness and the material you use, plus privacy basics for email and tracking. Seventh, a pre-publish checklist and an evidence file. The quest in this module, "claims check", practises the first half on an invented ad.
A note on what this is and is not. Rules differ by country, region and industry, and they change. This is general guidance to help you spot problems and ask the right questions. It is not legal advice. For anything high-stakes, such as health, finance, children's products, regulated industries or a large campaign, check the official guidance for where you sell, or ask a qualified professional.
You will finish with a claim-testing method, a list of high-risk areas, rules for testimonials and endorsements, a labelling guide and a one-page checklist.
Who This Is Useful For
What You Will Learn
What You Need
1. Marketing material to review: an ad, a landing page, an email, a social post, or your current best-selling product page.
2. The facts you have: product details, test results, reviews, prices, certificates.
3. A way to find the rules where you sell. Many regulators publish plain-language guidance for businesses.
4. A chat AI.
5. About ninety minutes the first time, then a few minutes each time you publish something new.
Information as of 14 November 2026. Laws, platform policies and enforcement priorities change often, and they differ by country, region and industry. This tutorial describes common principles, names no specific legal thresholds and gives no guarantee that anything complies with your local rules. Where it mentions the EU's transparency rules or US endorsement guidance, treat that as a pointer to look up the current official text. Do not paste customers' personal data into an AI tool.
Step 1 — The claim test
A claim is any statement in your marketing about what your product is, does or achieves, or about you, your customers or your competitors. It can be words, numbers, images, a before-and-after photo, a rating badge or an implication. Most trouble comes from claims no one thought of as claims.
Run every claim through five questions.
1. What exactly does it say, and what would a customer take it to mean? The test is the overall impression on a reasonable customer, not your intention or the small print. "Our cream is the number one choice" may be taken to mean best-selling, best-rated or most recommended by experts.
2. Can I prove it, with evidence I hold now? Proof means something you could show a sceptical customer, a platform or a regulator: test results, records, receipts, a study, a certificate, a document. "We believe" or "everyone knows" is not proof. You generally need to have the evidence before you make the claim, not after someone asks.
3. Is the proof good enough for the strength of the claim? The bigger and more specific the claim, the stronger the proof needed. "Customers love it" needs less than "reduces back pain in two weeks".
4. Is anything missing that changes the meaning? A true statement can mislead if it leaves out an important condition. "Free delivery" while charging for the product at a higher price, or "30-day guarantee" that excludes most cases, can mislead by omission.
5. Is it about something regulated? Health, medicine, food and supplements, money and investments, children, safety, environmental benefits and some professional services face stricter rules.
Then act on the answer. For each claim, choose one:
| If | Then |
|---|---|
| You can prove it and it is clear | Keep it, and file the evidence |
| It is true but could mislead | Add the missing condition, or reword it |
| It is probably true but you cannot prove it | Gather evidence first, or soften to what you can show |
| You cannot prove it | Remove it |
| It touches a regulated area | Check the rules for your market before using it |
Opinion and puffery. Loose, obviously subjective praise ("the tastiest tea you will try") is often treated differently from factual claims ("the only tea with no added sugar"). But the line is not always clear, and platforms may still reject superlatives. When a statement can be tested, treat it as a factual claim.
Step 2 — The claims that cause the most trouble
Some kinds of claim attract complaints and regulators. Know them, and be extra careful.
"Best", "number one", "leading", "most trusted". Superlatives imply you have compared yourself with others. If you cannot show who, how and when, drop them or specify: "rated 4.7 from 61 customer reviews".
"Clinically proven", "scientifically tested", "doctor recommended". These imply specific, credible evidence. Use only if you have the actual studies or endorsements, they apply to your product and the exact wording is supported. A study of an ingredient is not a study of your product.
Health, medical and wellbeing claims. Claims that a product treats, prevents or cures a condition, helps weight loss or improves health are among the most regulated. Avoid them unless you have strong evidence and know the rules for your market and product type.
Results and earnings claims. "Make HK$50,000 in your first month" or "lose 5 kg in two weeks". If you show a result, it must be typical, or you must say what is typical. A single exceptional case presented as the norm misleads.
Before-and-after images. They imply your product caused the difference. They must be real, unedited, from the same conditions and representative.
"Natural", "organic", "eco-friendly", "sustainable". Environmental and ingredient claims are heavily scrutinised. Be specific and provable ("packaging made of 100 per cent recycled card") rather than vague ("green"). "Natural" with a synthetic ingredient on the label is a problem.
Comparisons with competitors. You may compare, but facts must be accurate, current, like-for-like and provable. Do not disparage, and be careful with other companies' trademarks and logos.
"Guaranteed", "risk-free", "no strings". Guarantees must be honoured, and conditions must be clear and not buried.
"Free". If a customer must pay something, such as delivery, a subscription after a trial or a purchase, say so clearly next to the word.
Customer numbers and ratings. "Used by 50,000 customers" or "rated 4.9" must be true, current and from a real source. A stale or inflated counter is a claim.
Authority claims. Awards, certifications, press mentions, "as seen in" and partner logos. Use only those you really have, and with permission.
Implied claims. Images, names and context can imply things: a white coat suggests medical approval, a government-style badge suggests an official stamp. Check what the design implies.
Fix by making it specific. The best repair for a doubtful claim is usually a precise, true one. "Best tea in Hong Kong" becomes "Teas from two small farms the owner has visited. About 60 per cent of buyers order again."
Step 3 — Testimonials, reviews and endorsements
Social proof is powerful, which is exactly why it is regulated.
Reviews and testimonials must be real. Never write, buy, swap or invent reviews. Do not ask AI to "write five customer testimonials". Do not create fake customer profiles or photos. In many places, fake reviews are expressly prohibited and enforced.
Use real ones accurately.
Incentivised reviews. If you give a discount, gift or payment in exchange for a review, many rules require clear disclosure, and some platforms forbid it altogether. Ask for honest reviews without conditions on their content.
Paid endorsements and influencers. If someone is paid, given free products, given commission or has another connection with you, and promotes your product, the audience is usually entitled to know.
Expert and "customer" personas. Do not present an actor, a stock photo or an AI character as a real customer or expert. If you use illustrative figures, make it clear they are illustrations.
AI-generated reviews and people. Do not use AI to create reviews, testimonials, "user" photos or video presenters that look like real customers. Clearly label any AI-generated spokesperson, and do not imply a real person endorses you when they do not.
Your own claims about reviews. "Rated 4.7 from 61 reviews" is a factual claim. Make sure it is true, from a source you can show, and updated.
Step 4 — Urgency, prices and offers
Pressure tactics and price games are among the most common sources of complaints.
Fake urgency.
If scarcity or a deadline is real, say it plainly and truthfully, with the real date or number.
Price claims.
"Free" and trials.
Bait and switch. Do not advertise something you do not intend to supply in reasonable quantity, or steer customers to a dearer item.
Prizes and giveaways. Competitions and giveaways have their own rules, and some places restrict them. State the rules, the odds or selection method, who can enter and what the prize is. Follow the platform's promotion rules.
Terms and conditions. Make important conditions visible where the offer is made. A condition that changes the meaning of the headline must not live only in a footnote.
Subscriptions and renewals. Make signing up and cancelling equally easy. Send clear reminders where required, and honour cancellations promptly.
Step 5 — Disclosure and AI labels
Honesty also covers who is speaking and how the content was made.
Disclose commercial relationships. Ads must be recognisable as ads. Paid posts, affiliate links, sponsored content and gifted products should be labelled clearly. If your blog earns commission on links, say so near the links.
Do not disguise ads as something else. An advertorial that looks like a news article, or a "review" that is really a paid placement, should be clearly marked.
Label AI-generated and AI-altered content where required or where it would mislead.
Product images must be accurate. An AI-generated image of your product must show the real product. If it shows a different colour, size, packaging, feature or result, it misleads, whether or not it is labelled. Use real photos for what the customer will receive, and clearly illustrative images only where they cannot mislead.
Do not misrepresent people. Do not use anyone's face, voice or name to suggest endorsement without permission. Do not impersonate real people, and do not create AI avatars of real people without their consent.
Be honest about who you are. Use your real business name and contact details. Do not hide your identity behind fake sender names or fake local presence.
Chatbots and automation. If customers interact with an AI assistant on your site or in messages, make clear that it is automated, and give a way to reach a person. Do not let a bot make promises you cannot keep.
Be transparent where it matters, not performative everywhere. You do not need a label on every spell-checked caption. The aim is to avoid misleading people in ways that matter to their decisions.
Step 6 — Copyright, likeness, and privacy basics
The material you use must be yours to use. Mistakes here are common, and often accidental.
Copyright.
AI-generated material.
Likeness and voice.
Privacy, email and tracking.
Regulated industries and categories. Health, finance, gambling, alcohol, tobacco, adult content, political advertising and children's products often have tight rules and platform restrictions. If you operate there, learn the rules before you create anything.
Step 7 — A pre-publish checklist, and an evidence file
Honest marketing becomes easy when it is routine. Run the same checklist before anything goes live, and keep the proof.
Use AI as a claims checker, not an approver. AI cannot tell you whether something is lawful, and it cannot know your evidence. But it is useful for spotting claims you did not notice and asking questions.
Role: a careful marketing compliance reviewer who is cautious and does not give legal advice.
Task: Read the marketing text below and list every claim it makes or implies (facts, numbers, comparisons, results, guarantees, urgency, price claims, endorsements, health or environmental claims). For each, tell me: (1) what a customer might take it to mean, (2) what proof I would need to hold to support it, (3) whether it could mislead by omission, and (4) a safer, more specific wording. Flag anything that touches a regulated area (health, finance, children, safety). Do not tell me whether it is legal. Remind me to check the rules where I sell.
[paste the text]Then you decide what is true, supported and acceptable.
The pre-publish checklist.
HONEST MARKETING CHECKLIST
Claims
[ ] I listed every claim, including numbers, images and implied claims.
[ ] I hold evidence for each, and I know where it is kept.
[ ] No "best", "number one", "clinically proven", "guaranteed" or similar without proof.
[ ] No health, money or safety claims unless supported and allowed in my market.
[ ] Results and before-and-after examples are real and typical, or I say what is typical.
[ ] Nothing important is missing that would change the meaning (conditions, fees, exclusions).
Reviews and endorsements
[ ] Every review and testimonial is real, accurate, permitted and current.
[ ] No invented, purchased or AI-written reviews, customers or experts.
[ ] Paid, gifted or commissioned promotion is clearly disclosed.
[ ] Influencers have a written brief of what they can claim.
Urgency, prices and offers
[ ] No fake countdowns, stock counts or deadlines.
[ ] "Was" prices are genuine. "Up to" and "from" prices are fair.
[ ] The full price and any conditions are clear before purchase.
[ ] Trials and subscriptions state price, renewal and how to cancel.
Disclosure and AI
[ ] Ads and sponsored content are recognisable as such.
[ ] AI-generated or altered content is labelled where required or where it would mislead.
[ ] Product images show the real product.
[ ] No real person's face, voice or name used without permission. No fake spokespeople presented as real.
[ ] Chatbots are identified as automated.
Rights and privacy
[ ] Images, music, fonts and text are mine or properly licensed.
[ ] No other brand's logo or content used to imply a link or imitate.
[ ] I have permission for any identifiable person.
[ ] Email consent, unsubscribe and tracking consent are handled. No customer data pasted into AI tools.
Final
[ ] I checked the rules and platform policies for where I advertise.
[ ] A second person has read it, for anything high-stakes.
[ ] Evidence and approvals are saved in my evidence file.Keep an evidence file. A simple folder or spreadsheet with: each claim, the evidence, the date and who checked it. Add review permissions, endorser agreements, licences for images and music, test results and screenshots of what was published and when. If a customer, platform or regulator ever asks, you can answer calmly and fast.
Review regularly. Claims go stale: prices change, ratings move, a study is superseded. Re-check live material every quarter and after any change to the product.
Handle complaints well. If someone says your marketing misled them, respond promptly and politely, investigate honestly, correct or remove the claim if it was wrong and, where appropriate, make it right. Treat a complaint as free feedback on your checking process.
Ask for help when it matters. For a big campaign, a regulated product or a gray area, get a professional opinion before you spend. It costs far less than a dispute.
Worked Example: cleaning up a serum ad
An invented skincare brand, GlowDrop, has this ad copy, drafted quickly with AI: "GlowDrop is clinically proven to erase wrinkles in 7 days. Used by 50,000 happy customers, recommended by dermatologists, made with 100% natural ingredients. Only 12 left, sale ends tonight! Try it with our 30-day money-back guarantee." The owner runs the checklist, and also uses the AI reviewer prompt.
The claims and what she finds.
1. "Clinically proven to erase wrinkles in 7 days." She has no study of the product. A supplier's note about one ingredient is not proof of this result, and "erase wrinkles" is a strong cosmetic and arguably medical-style claim. She removes it. If she wants to say something, she can describe what she can support, for example "a lightweight serum with hyaluronic acid", with the ingredient list.
2. "Used by 50,000 happy customers." Her order records show about 1,900 orders. The number came from an old counter. She replaces it with the true figure: "More than 1,800 customers", with the date she checked.
3. "Recommended by dermatologists." The only recommendation is from a paid brand ambassador who is a dermatologist, and the page does not say so. She removes the generic claim. If she keeps the quote, she will attribute it, and clearly disclose the paid relationship.
4. "100% natural ingredients." The ingredient list includes a synthetic preservative. She removes "100% natural" and says specifically "No added fragrance. Contains a plant-derived hyaluronic acid", checking each point against the ingredient list.
5. "Only 12 left, sale ends tonight!" The timer resets and stock is plentiful. She removes both. The sale is real for the month, so she says: "Launch price until 30 November", and means it.
6. "30-day money-back guarantee." The refund policy page states the same 30 days and how to claim, and she honours it. She keeps this claim, and links the policy next to it. This one has real evidence.
The imagery. The AI-generated "before and after" images show smoother skin than her product delivers. She removes them and replaces them with real, unretouched photos from consenting customers, with permission on file, and a note that results vary.
The rest. She adds the full price and delivery cost near the button, confirms the music in her video video is licensed, and labels a synthetic voice-over in the video as AI-generated. She saves the evidence in a folder: the order records, the ingredient list, the refund policy, the permissions and the licence receipt.
The revised ad: "GlowDrop: a lightweight serum with hyaluronic acid and no added fragrance. More than 1,800 customers (checked 14 November). Launch price until 30 November. 30-day money-back guarantee." It is plainer, and every line can be backed up.
Common Mistakes to Avoid
Three habits cause most advertising problems.
Mistake #1: Letting AI or habit write claims you cannot prove. Fluent text ("clinically proven", "thousands of customers", "best") slips into ads unnoticed, and customers and regulators judge what you published. Fix: list every claim, require evidence for each, rewrite what you cannot support in specific true terms and keep an evidence file.
Mistake #2: Faking proof or pressure. Invented reviews, AI "customers", resetting timers and inflated "was" prices might raise short-term sales, and they damage trust and can break the rules. Fix: use only real, permitted reviews, disclose paid relationships, state real limits plainly and keep prices consistent.
Mistake #3: Using material you have no right to use, or hiding what it is. Borrowed images and music, someone's face or voice, or realistic AI content that looks real and is not labelled, cause disputes and takedowns. Fix: use your own or properly licensed material, get permission for people, label AI-generated content where required or where it would mislead and make sure product images show the real product.
Going Further
Write a one-page claims policy. A short list of what your business will and will not say, such as no health claims, no "best" without proof and no fake urgency. Paste it into every AI prompt that writes copy, and share it with freelancers and influencers.
Build an approved-claims library. A list of true, supported statements with their evidence, ready to reuse. It makes writing faster and safer.
Learn the rules where you sell. Find your regulator's plain-language guidance for advertisers. Many publish short guides and checklists for small businesses.
Train anyone who writes for you. A short briefing on the checklist saves much more than it costs.
What to learn next. This completes the AI for Marketing track. A good next step is to take one live campaign and run the honest-marketing checklist and the weekly report on it for a month. The "claims check" quest lets you practise spotting unsupported claims on an invented ad.
Key Takeaways
In this tutorial you learned:
Before you move on, a challenge. Take one live ad or page, and list every claim in a table with its evidence. Rewrite or remove any you cannot support, run the pre-publish checklist and save an evidence file. Then do the "claims check" quest to practise on an invented ad.