AI is rapidly becoming an ordinary production tool for advertising. Copy, product images, voice, video, translations, backgrounds, and campaign variations can all be generated or modified with AI.
That creates an awkward question for brands: when should the audience be told?
Google moved further toward disclosure in July 2026 by introducing a “How this ad was made” section in My Ad Center. Ads created with Google's own generative tools can be identified automatically, while advertisers can indicate when outside AI tools were used. The Google announcement reflects a wider shift toward making synthetic media easier to identify.
In Europe, the issue is no longer only voluntary. AI Act transparency obligations began applying on August 2, 2026, including requirements around machine-readable marking and disclosure for specified kinds of AI-generated or manipulated content. The European Commission guidance makes clear that transparency rules are becoming part of the operating environment.
But “AI was involved” still covers an enormous range of activity.
A label can be useful and meaningless at the same time
Imagine three advertisements.
One uses AI to remove a background from a real product photograph. Another creates a fictional model wearing a real product. A third generates a fake customer testimonial from a person who never existed.
Calling all three “AI-generated” treats fundamentally different situations as equivalent.
That is the weakness of blanket labeling. AI is becoming embedded in editing tools, cameras, design software, ad platforms, and marketing workflows. Eventually, asking whether AI touched an advertisement may become as broad as asking whether Photoshop touched it.
The more important question is whether AI changed something a reasonable customer would consider material.
Disclosure should follow the risk of deception
Marketing already has a useful principle: do not mislead people about what they are buying.
AI makes that principle harder to apply, but it does not make it obsolete.
If generative tools create a fictional lifestyle scene around a real bottle of shampoo, a visible disclosure may add little. If they generate a spokesperson who appears to be a real doctor, fabricate a product demonstration, invent a customer, alter a before-and-after image, or simulate an event that never happened, disclosure becomes much more meaningful.
Brands that already think carefully about omnichannel marketing should apply the same consistency to AI provenance. A customer should not receive one level of transparency on a website and another in a paid social advertisement simply because different production tools were used.
Trust may become a competitive advantage
There is a fear that AI labels reduce performance because they remind people that an image or message is synthetic. That may be true in some contexts.
The opposite effect is also plausible. As synthetic media becomes more realistic, brands that explain how content was created may appear more trustworthy than brands that force audiences to guess.
This is particularly relevant when building long-term brand visibility in AI and search. Trust signals increasingly matter across channels because customers encounter a brand through search results, AI answers, social ads, creator content, and automated campaigns before they ever speak with the company.
Transparency can become part of that brand system.
Not every AI edit needs a warning label
A useful policy is to disclose AI when the synthetic element could materially change what the customer believes.
That is stricter than saying disclosure is never necessary and more practical than labeling every resized image or rewritten headline.
Brands should also keep an internal record of how major creative assets were produced. Regulation will evolve, platforms will change their disclosure rules, and customers may ask questions later. Knowing what was generated, what was altered, and what was real is becoming basic marketing governance.
The strongest brands will probably use AI heavily. They will also understand that efficiency does not remove responsibility.
The question is therefore less “Did AI make this ad?” and more “Would a reasonable person interpret this ad differently if they knew how it was made?”
When the answer is yes, tell them.
Originally published on the Mustard Seed blog.
