AI Disclosure Rules Every Creative and Consulting Pro Should Know

Why disclosure matters more than the AI itself

Marketing agencies, photographers, e-commerce sellers, consultants, and coaches have all found the same thing over the past couple of years: AI tools can multiply output. Draft copy, mockups, pitch decks, retouched photos, research summaries, and client reports can all move faster with AI in the loop.

What hasn’t changed is who is responsible for what gets delivered. If you put your name on it, sell it, or use it to make a claim to a client or the public, you own the accuracy and the honesty of that work, regardless of what tool produced the first draft. The businesses that get into trouble aren’t the ones using AI. They’re the ones treating AI output as if it needs no review, no disclosure, and no fact-check before it goes out the door.

The FTC rules that actually apply to creative and advisory work

The Federal Trade Commission has long-standing rules on endorsements, testimonials, and deceptive claims. These rules don’t have a special exemption for AI. If anything, AI-generated content gets more scrutiny because it’s easier to produce something that looks like a real result without being one.

Fake reviews and testimonials

Generating a “client testimonial” with AI, even as a placeholder, and using it in real marketing material is a direct violation of endorsement guidelines. This includes:

• AI-written reviews attributed to real or fictional customers
• Composite testimonials that blend outcomes from multiple clients into one quote
• Star ratings or review counts that aren’t tied to actual verified feedback

If you use AI to help draft a request for a review, or to summarize real feedback into a shorter quote, that’s fine as long as the underlying sentiment is genuine and the client approved the final wording attributed to them.

Results and outcome claims

“Clients see a 3x return within 90 days” is a claim that needs to be backed by actual data you can produce if asked. AI is very good at generating plausible-sounding numbers, projections, and case study language. None of that is usable in marketing unless it reflects real, documented outcomes from real client work.

A simple internal rule helps here: any number in outward-facing material needs a source you could point to in five seconds. If you can’t find the source, the number doesn’t go out.

Endorsements by AI personas

If you’re using an AI-generated spokesperson, avatar, or voice in ads or social content, that’s an emerging area of scrutiny. Audiences generally assume a person speaking in a testimonial-style format is a real customer or real endorser unless told otherwise. Label AI-generated presenters clearly.

Disclosure: when, where, and how much

There’s no single federal law mandating a specific disclosure label for all AI-assisted content, but disclosure obligations exist in several overlapping places: FTC deception rules, platform policies (many ad platforms and social networks now require AI content labels), and increasingly, state laws around synthetic media.

A practical disclosure framework

Rather than trying to track every jurisdiction’s exact wording requirement, use a tiered approach based on how the content is used:

• Internal drafts and brainstorming: No disclosure needed. This is a tool, like spellcheck or a stock photo library.
• Client deliverables: Disclose in your contract or scope of work that AI tools are used in your process, generally. You don’t need a label on every asset, but the client should know upfront, not discover it later.
• Public-facing content with factual claims: Disclose specifically if the content includes AI-generated images, video, or data visualizations that could be mistaken for real photography, real footage, or real research.
• Reviews, testimonials, or endorsements: Never use AI to generate these. This is the one category where disclosure isn’t enough. It shouldn’t be done at all if it’s not real.

Where to put the disclosure

For images and video, a small persistent label (“AI-generated” or “AI-assisted”) in a corner or caption is standard practice and increasingly expected by platforms. For written content, a line at the end of a blog post or report noting that AI tools were used in research or drafting is enough for most contexts. For ads specifically, check the platform’s current policy since these are being updated frequently and some now require metadata tagging, not just a visual label.

Copyright and licensing: what you actually own

This is where a lot of confusion lives. A few practical points:

• Purely AI-generated output, with no meaningful human creative input, generally cannot be copyrighted in the US. If a client is paying for exclusive ownership of creative work, pure AI output without substantial human editing and direction may not qualify for the protection they think they’re getting.
• Work that combines AI-generated elements with substantial human creative decisions, editing, arrangement, and refinement has a stronger claim to copyright protection, though the exact line is still being worked out in case law.
• Check the terms of service of whatever AI tool you’re using. Some tools claim broader rights over inputs or outputs than users expect. Read the commercial use section specifically before using outputs in paid client work.
• If you’re using AI on licensed stock assets, client-supplied photos, or copyrighted reference material, make sure your license actually permits AI-based transformation. Not all stock licenses do.

Protecting yourself contractually

Put a short AI-use clause in client contracts covering three things: that AI tools may be used in the delivery process, who owns the final work product, and what happens if a platform or law later requires specific disclosure that wasn’t anticipated at signing. This is a few sentences, not a rewrite of your whole contract, but it closes a gap that’s increasingly coming up in disputes.

Building this into your actual workflow

The businesses handling this well tend to do a few concrete things:

• Keep a simple internal log of which deliverables involved significant AI generation, so you can answer questions quickly if a client or platform asks
• Have one person, even in a small shop, responsible for a final “is this true and is this disclosed correctly” pass before anything client-facing or public ships
• Review platform ad policies quarterly since AI content rules are changing faster than most other ad policy areas right now
• Never let AI draft anything that will be presented as a direct quote, review, or first-person account from a real person without that person’s explicit review and approval

None of this requires slowing down your actual production process much. It mostly requires deciding, in advance, where the line is between “AI helped me work faster” and “AI is being passed off as something it isn’t.” Get that line clear once, write it down, and the day-to-day decisions get a lot easier.

For the complete, structured playbook on this topic, see AI for Marketing, Creative & Consulting in our library. New here? Start with our free guide.