When Consumers Recognize AI: Changing Patterns in AI-Generated Marketing

It would be foolish not to leverage the efficiency gains of AI tools for marketing campaigns. Yet, as research shows, the performance of AI-made ads, and by extension other marketing content, is reduced if AI use is disclosed. But the same research also found that fully AI-generated ads can raise click-through rates in other settings.

The creative doesn’t change, only the label signifying its origin and shaping viewer perception. With the direction of EU AI transparency rules, more marketing materials will likely be labeled as AI-generated in the future. Since performance differences stem from the perceived origin of creatives, marketers should consider treating AI use as a positioning decision. This shift is taking place against a rapidly expanding market: the generative AI in advertising market is estimated at USD 4.18 billion in 2026 and, assuming a 23.8% CAGR, could reach approximately USD 19.91 billion by 2033. As generative AI moves from experimentation into everyday advertising workflows, the question is no longer simply whether marketers will use it, but how consumers will respond when they recognize it.

Perceived Origin of AI

Three types of ads were examined for a real-world beauty e-commerce shop: human expert-created ads, genAI-modified ads (genAI enhances expert designs), and genAI-created ads (generated entirely by visual genAI). The ads were then used in the Google Display Network in the United States to examine real-world results.

Fully generated ads showed an increase in click-through rates, while modified ads showed no significant improvement over the human benchmark. The effectiveness also increased when AI was given greater freedom to redesign the product packaging. The conditions differed in how much room the AI tool was given, and so did the results.

That connection between AI production and advertising outcomes is particularly relevant to the application side of the market. Content generation is estimated to account for approximately 32% of the generative AI in advertising market in 2026, reflecting the growing role of AI in producing the very creative assets, copy, visuals, and campaign material that consumers increasingly encounter. The wider application landscape includes Content Generation, Personalized Advertising, Customer Engagement, Market Analysis, and Other Applications.

A larger dataset in another study points the same way. This study compared over four thousand ads launched by the same advertisers in identical campaign settings at the same time. Across many impressions, they found no detectable average click-through disadvantage for the AI images.

Controlled-environment studies that disclosed AI involvement in ad generation found that advertising effectiveness was reduced.

The findings are further complicated by the fact that people evaluators are rather bad at identifying AI-made ads.

The results suggest that the perceived AI origin of the ad is the reason behind performance differences. Since humans aren’t that good at differentiating ads, any AI label the creative or even the brand is associated with will result in lower performance.

The best strategy, it seems, would be to use AI creatives but make it appear fully human-made.

How Far the AI Label Reaches

The strategy of using AI creatives and framing them as human raises clear compliance issues. Currently, the most far-reaching legislation, the EU AI Act, imposes a visible label requirement for deepfakes and AI-generated text on matters of public interest.

A generated product shot, such as a model for an online beauty retailer, looks like neither, and the Act doesn’t require a badge. Yet to know for sure, we must apply the criteria of deceptive similarity required by the AI Act's transparency obligations.

If the content resembles something that plausibly exists and would falsely appear authentic, it is a deepfake. The last point is judged against what the intended audience expects. The AI model might not exist, but if it is realistic enough, it could. So do the room behind, the lighting, and every other detail used.

The line between AI-generated content and deepfakes is easy to cross, and mistakes can cost a lot. If a creative is reported or authorities disagree, a fine can reach millions. Of course, this is EU legislation, but one can expect other countries, including the United States, to follow with similar methods of controlling AI use.

That connection between AI production and advertising outcomes is particularly relevant to the application side of the market. Content generation is estimated to account for approximately 32% of the generative AI in advertising market in 2026, reflecting the growing role of AI in producing the very creative assets, copy, visuals, and campaign material that consumers increasingly encounter. The wider application landscape includes Content Generation, Personalized Advertising, Customer Engagement, Market Analysis, and Other Applications.

The EU legislation already requires AI providers to mark the output in a machine-readable form. The marking is in the file before the image is edited or uploaded to a platform, but since it’s only meant for automated systems, people won’t see it. If the ad crosses the deepfake line, they responsible for the visible label. This becomes particularly relevant in cross-platform campaigns, where the same AI-generated creative may be adapted and distributed across different channels as part of a broader Cross Platform and Mobile Advertising strategy.

Many prominent platforms already read content credentials and apply labels without asking the advertiser, using the C2PA standard that predates the EU Act. Even if no label ever appears, the strategy still asks the team for something specific. Take a generated ad and make it read as if a human made it.

AI Use as Brand Positioning

Plenty of marketers already use a workflow of generating ads and then backtracking the parts that give away their AI origins. The studies we discussed earlier show that human design around AI inputs doesn’t produce additional gains, while removing constraints can improve performance.

Editing away the less realistic qualities the model produced is the condition research ties to the performance increases. So, the strategy lowers the perception at the cost of performance.

However, it remains unclear how much of the performance improvement survives human edits, while constrained AI ads did not perform better than human ads. You’re paying for an outcome you cannot guarantee, since the ad platform might decide for you.

The evidence is scarce and based on click-through rate and purchase-intent impacts spanning only a few months. In the long run, results for a brand that heavily uses AI and hides it might be more negative. It also varies by industry, specific products, and other brand-positioning factors.

The industry context matters here as well. Retail is estimated to represent approximately 27% of the generative AI in advertising market in 2026, making it especially relevant to the discussion of AI-generated product imagery, e-commerce campaigns, and consumer-facing advertising. Other industry verticals include Healthcare, Financial Services, Automotive, and Others, each presenting different expectations around personalization, authenticity, compliance, and brand communication.

Click-through and purchase intent measure months, while the choice about AI use in creatives plays out over years. The available data can't settle it, so it may be best to treat AI use in marketing as a long-term positioning decision aligned with what the brand stands for across campaigns.

Currently, three strategies for using AI in marketing ads and other content seem most plausible and readily available.​

• Suppress. Keep AI generation constrained to the ideation stage, and avoid the risks of creating deepfakes or having your campaigns labeled AI-generated. It’s the most defensible position for brands where craft is part of what’s sold and can afford to lose the AI generation gains entirely.
• Disclose and absorb. Label AI use as a normal routine, accept a short-term hit on metrics, and hope performance stabilizes once the AI label is normalized. Best for brands that compete on the product rather than how the ad was made.
• Own it. Make AI-native production a part of the brand's identity. It’s the highest risk of the three, but also the only position that gains if judgements of AI creative soften as the practice becomes ordinary.

Any of the three strategies can work for different brands. What’s perhaps the worst approach is lacking consistency in adopting one of them. Treating it as a brand positioning problem will allow you to change the stance if the public's perception shifts.

The competitive landscape is developing alongside these changes, with companies such as Meta Platforms Inc., Amazon Web Services Inc., Accenture plc, IBM Corporation, NVIDIA Corporation, Baidu Inc., Adobe Inc., Merkle Inc., Weber Shandwick International Ltd., HubSpot Inc., Zeta Global Holdings Corp., DataRobot Inc., Hootsuite Inc., BrightEdge Technologies Inc., Persado Inc., Matellio Inc., Brafton Inc., Bannerflow AB, ClarifAI Inc., and Aegis Softech participating across AI, advertising technology, marketing platforms, analytics, and related services.

Conclusion

Marketing metrics show that origin and perceived origin of creatives influence performance in opposite directions, so AI use should be treated as a brand positioning decision. Legislation and detection will keep evolving, and penalties may soften as generated creatives become ordinary. What can be controlled by marketers is their brand positioning. For brands operating in the U.S. generative AI in advertising market, the challenge will be balancing the efficiency and creative possibilities of AI with transparency, consumer expectations, and a consistent identity across campaigns.