AI Content Disclosure: What to Tell Customers

Howard Gossage, the great San Francisco copywriter, wrote in 1961 that the public was not stupid, that the public was in fact much smarter than most advertising assumed, and that advertisers got into trouble whenever they forgot it.

That observation has aged into something close to a law. It is also, in 2026, the most useful frame for thinking about AI content disclosure.

Most brand teams we work with come to AI content production with a substantial amount of anxiety about disclosure. The standard question, asked early in every engagement, is some version of: do we have to tell customers? What if they find out? What if they object?

After running AI production for a few dozen brands over the last eighteen months, our short answer to this is that brand teams are worrying about the wrong question. The actual question is upstream, and it is simpler.

The question that actually matters

Audiences do not have a problem with AI-augmented brand content. They have a problem with brands that misrepresent reality.

These are not the same thing.

A brand that shoots its hero product, generates lifestyle context through AI, and posts the result on Instagram is not misrepresenting anything. The product is real. The brand is real. The context is implied — same as it has been in every catalogue shoot of the last hundred years. A skincare brand that photographs its serum on a white marble counter is not promising every customer a white marble counter. The audience knows this.

A brand that fabricates customer testimonials, invents before-and-after results, or shows product capabilities the product does not actually have, on the other hand, is misrepresenting reality. The technology used to fabricate is not the issue; the fabrication is. This was a problem before AI and remains one after it.

The line is between augmentation and fabrication. AI-augmented context, environment, scale, variation, and style — fine. AI-fabricated outcomes, capabilities, customer experiences, or proof — not fine. The audience can tell the difference, and so can regulators.

What our practice looks like

A few rules we apply, in increasing order of strictness.

One. Real product. Always. The product depicted is the actual product. The packaging is the actual packaging. The ingredient list, label, and physical features are accurate. We do not regenerate the bottle, retouch the label, or alter what is inside the package.

Two. Real people for people-driven content. Founder portraits, customer testimonials, team moments, before-and-after results from real customers — these are shot or sourced, never generated. The trust threshold is too high to play with.

Three. Honest claims. The brand's claims about what the product does are what the brand can substantiate. If the moisturiser does not reduce wrinkles by forty percent, the imagery should not imply that it does. This is unchanged from any other production context.

Four. Context and environment are augmentable. The white marble counter, the warm bathroom light, the seaside backdrop, the ingredient close-up, the lifestyle scene — all of this is environment, all of this is implied, all of this is fine to generate.

Five. Stylised concept work is augmentable. Abstract scenes, atmospheric video, conceptual imagery that does not claim to be photographically real — fine.

Six. Variation and scale are augmentable. The same product in twelve different contexts, the same brand in twelve seasonal variations, the same hero image at twelve aspect ratios — this is the highest-leverage and most policy-clean use of AI.

The brands we work with that operate by these rules have not had a single customer complaint, regulator inquiry, or platform takedown attributable to AI production. The rules are not novel; they are extensions of the advertising disciplines that have always applied.

What about explicit disclosure?

Brand teams often ask whether the imagery itself should be labelled — "AI-generated" badge, footnote, asterisk.

Our view, in current practice: usually no, sometimes yes.

In commercial brand and ad imagery — the listing photo, the Instagram post, the ad creative — explicit labelling is not currently required by any major platform or regulator we operate in, and labelling tends to draw attention to the production method rather than the product. We do not recommend it for typical brand content.

In editorial or journalistic contexts — a piece presented as documentary, a customer story, a behind-the-scenes account — explicit labelling is appropriate. The framing matters. Content presented as a record of reality should be a record of reality.

In political, financial, or health-claim contexts — where AI disclosure rules are tightening rapidly across jurisdictions — explicit labelling is required and our practice is to lean into transparency. The downside of under-disclosing in these contexts is large; the downside of over-disclosing is negligible.

In any context where the brand is making outcome claims — weight loss, hair growth, skin condition improvement — the imagery used to substantiate the claim should be unambiguously real. AI in these contexts is risky regardless of disclosure.

What audiences actually think

The empirical question — what do audiences actually feel about AI-augmented brand content — is more boring than the discourse suggests.

Audiences who are paying close attention to AI tend to notice and care; audiences who are not, do not. The vast majority of consumers seeing brand content on Instagram, Amazon, and Meta ads are not consciously evaluating whether the image was AI-augmented. They are evaluating whether the brand looks good, whether the product looks like something they want, and whether the price is reasonable.

When audiences are explicitly asked — in surveys and focus groups — they consistently report that they care more about whether the product matches what arrives in the box than about whether the imagery was AI-augmented. Brand trust is downstream of product experience. Production method is, for most consumers, an abstract concern.

The exception is when a brand misuses AI in ways that violate trust. Fabricated testimonials, false outcomes, generated celebrity endorsements without permission — these damage brand trust acutely, and the damage compounds when discovered. The lesson is not that AI is dangerous. The lesson is that misrepresentation is dangerous, and AI is a particularly efficient way to do it if you choose to.

The competitive consequence of getting this right

Brands that adopt disciplined AI content production tend to find that disclosure, when it comes up, is a non-issue. Customers ask occasionally; brand teams answer honestly; the conversation ends.

Brands that adopt AI production without discipline — that generate fabricated content, that misrepresent the product, that try to optimise for engagement at the cost of trust — eventually face a much larger problem, and it is not a disclosure problem. It is a brand-equity problem and, increasingly, a regulatory one.

The brands that get this right today will be operating with a content pipeline that compounds for years. The brands that get it wrong will spend years rebuilding trust. The cost of getting it right is mostly the cost of discipline. The cost of getting it wrong is much larger.

What we would suggest

If you are evaluating an AI content production partner, the questions to ask are about discipline, not just capability.

Ask what they will not generate. The list should be specific — faces, customer outcomes, fabricated testimonials, regulated claims — and it should be operational, not theoretical.

Ask to see how they handle disclosure when a customer asks. The brand team will get questions; they should know what to say.

Ask about their position on AI in regulated categories — supplements with health claims, beauty with active ingredient claims, financial services, political. A partner without a clear position on these is a partner without discipline.

If you want our position written down for your reference, our AI Creative Production practice page lays out the four operating principles we apply across every brand we work with.

Frequently asked

Do platforms require AI disclosure? For brand content, currently no on Meta, Google, Amazon, and most major platforms. For political and "issue" content, yes on Meta and Google. For specific regulated content (health claims, financial), the rules are tightening and vary by jurisdiction. We monitor monthly.

What does the law say in India? Indian advertising law applies to AI-augmented content identically to other commercial content. The ASCI guidelines require ads to be truthful, not misleading, and substantiable. AI does not change these requirements.

What about EU and UK regulations? EU AI Act and UK regulations require disclosure for AI-generated content in specific categories — deepfakes, political content, content that could be mistaken for human-generated journalism. Standard brand and product content is not generally required to be labelled.

Will the rules change? Yes. Platform policies and regulations are tightening across jurisdictions. The direction of travel is toward more disclosure for politically and health-sensitive content, not toward labelling of standard brand imagery. We adjust our practice as the rules change.

What about US FTC guidelines? The FTC requires endorsements and testimonials to be genuine and substantiated. Generated testimonials, fake customer reviews, and fabricated endorsements are prohibited. This applies identically whether the fabrication is AI-driven or otherwise.

---

If you operate a brand and want a written AI content discipline policy tailored to your category and regulatory environment, write to us at connect@yatharthchopra.com. We do this as part of every engagement.