Disclosure and compliance for AI-generated marketing

Most of the anxiety about AI in marketing resolves into a question with a stable answer: does the audience end up with a false impression of something that matters to them. That test predates generative tools by many decades, and it still does most of the work.

What has changed is that the cost of producing convincing false impressions has collapsed, and that a franchise system has hundreds of people publishing in a shared brand name. Both make policy more urgent than it used to be.

A usable line: process versus substance

The practical distinction that resolves most cases is whether the AI involvement is a production method or part of what is being represented.

The left column is ordinary tooling and generally does not require disclosure, in the same way that nobody discloses the use of a spellchecker or a photo editor. The right column changes what the audience believes to be true, and that is where both disclosure obligations and deception exposure sit.

The test to apply to any borderline case: if the audience learned how this was made, would they feel misled about something they cared about. If yes, disclose it or do not make it.

The things not to do at all

Some uses are not a disclosure problem — a label does not fix them, because the content only works by being believed.

1. Synthetic testimonials or reviews. A generated customer endorsement is a fabricated endorsement. This is squarely within existing consumer protection enforcement and disclosure does not cure it
2. Generated images of locations, staff or work presented as real. A rendered storefront, a synthetic team photograph, a fabricated before-and-after
3. Cloned voices or likenesses without documented consent, including of your own franchisees and staff. Consent should be specific about use and duration, and it should be in writing
4. Generated results or performance claims. In a franchise context this collides directly with financial performance representation rules — if it is not in your Item 19 you cannot publish it, whoever or whatever produced it. See Item 19 and your recruitment marketing, and confirm the specifics with your franchise attorney
5. AI assistants presented as people, which is covered in AI chat in franchise recruitment

The first two are where franchise systems are most exposed, because the temptation is highest exactly where the need is greatest: a new location with no photographs, no reviews and no completed work has an obvious gap, and generated content fills it perfectly and disastrously.

Franchise-specific exposures

Content published in a franchisee's name

Centrally generated content that goes out under a local owner's name and profile creates an exposure for a business the franchisor does not own. If a generated post makes a claim that does not hold for that location — a service they do not offer, an availability they cannot meet, an implied result — the complaint attaches to them.

Decide explicitly who is responsible for approving local content before it publishes, and make sure the owner has a genuine opportunity to review rather than a theoretical one. A scheduled queue that publishes automatically unless the owner intervenes is not meaningful approval.

Franchisees generating their own

The larger exposure runs the other way. Owners have the same tools, less familiarity with advertising rules, and a strong incentive to produce material quickly. A franchisee generating a testimonial, a before-and-after, or an earnings claim creates a brand-level problem — and an earnings claim is one you cannot make even when an owner is the one making it, unless it sits inside your Item 19. Take the position your franchise attorney advises and give it to owners in writing.

This belongs in onboarding, in plain language, with examples — not in a policy document nobody reads. The pre-opening window described in franchisee onboarding is the right moment, because it is before the owner has developed habits and before the first thing gets published.

Recruitment material

Franchise recruitment marketing is subject to franchise disclosure regulation in addition to general advertising law. AI-generated content in recruitment therefore carries two sets of obligations at once, and it is the area where the review process should be strictest. A generated image of a thriving location used in recruitment material is a substantially worse problem than the same image used in a consumer ad.

Where disclosure is genuinely required

Specific obligations vary by jurisdiction and are expanding, which is why this needs a live answer from your franchise attorney rather than a static one. The categories where requirements are most commonly encountered:

• Automated systems interacting with consumers, such as chat assistants, where several jurisdictions require disclosure
• Synthetic depictions of real people, which attract both disclosure requirements and right-of-publicity considerations
• Political and issue advertising, which is largely irrelevant to franchise marketing but has driven much of the rule-making
• Platform-level requirements, which are separate from law — major advertising and social platforms have their own synthetic media disclosure policies, and breaching those costs you distribution regardless of legality
• Sector-specific rules, particularly anywhere health, financial services or claims about outcomes are involved

That fourth point is worth noting because it is the one most often missed. Platform policy is enforced faster than law and the consequence — account restriction — lands immediately on the campaigns you are running.

A policy that survives contact

Franchise systems need something short enough that a busy owner reads it and specific enough to be applied without judgement calls.

1. A list of things nobody may generate, stated concretely: customer testimonials, images of people presented as real customers or staff, images of locations or work, voices, results
2. A statement of what is fine, so the policy does not read as a prohibition on AI. Drafting, editing, translating, summarising, backgrounds, analysis
3. A named approver for local content and for anything used in recruitment
4. A disclosure standard for the cases that require it, with the exact wording provided so nobody has to compose it
5. A consent record for any use of a real person's likeness or voice, including franchisees and staff, held centrally
6. A review date. Twelve months at most, and sooner if the regulatory position moves

One page. A longer document will not be read by the people whose behaviour it is meant to govern, which is the entire purpose.

The reputational layer

Separately from what is required, there is what is wise. Audiences are becoming quick at spotting generated imagery, and a brand caught using a synthetic customer photograph loses something that no disclosure regime was protecting.

For franchise brands this matters more than for most, because local trust is the product. A location page whose warmth comes from generated stock people is competing against a competitor whose photographs are visibly of actual staff, and the comparison is not favourable — which is a large part of why an imperfect real video outperforms polished synthetic material, as argued in filming an owner introduction on a phone.

The defensible position, and the one that ages well: use AI heavily for production, drafting, analysis and scale, and keep every depiction of a person, a place or a result genuine. That line is easy to explain to franchisees, easy to audit, and unlikely to be made obsolete by whichever disclosure rule arrives next.

Questions we get asked