AI-Generated Ads Can Match Human Creativity — What the Latest Study Means for Performance Marketing

Artificial intelligence continues to reshape the digital marketing landscape, but one lingering question has persisted: Can AI-generated advertising truly perform as well as human-created campaigns in the real world? A major new academic study provides one of the most comprehensive answers yet. The research, conducted by a team from Columbia University, Harvard University, the Technical University of Munich and Carnegie Mellon University, analysed real live campaign data and found that AI-generated display ads can deliver click-through rates (CTR) on par with human-made ads — and sometimes even outperform them under the right creative conditions.

Breaking the Hype With Real-World Data

Much of the existing evidence about generative AI in advertising comes from lab experiments or small pilot tests. What makes this study especially noteworthy is that it used real market data. Researchers analysed more than 300,000 ads, totalling over 500 million impressions and about 3 million clicks, drawn from campaigns running on Taboola’s Realize performance marketing platform. This scale gives the findings stronger ecological validity than most previous research.

Rather than running a simple A/B test, the study paired “sibling ads” — matching an AI-generated creative with a human-made one from the same advertiser, running on the same day, with identical objectives and landing pages. This controlled design helped isolate the influence of creative origin itself, ensuring the comparisons were fair.

AI Keeps Pace With Human Creatives

Across this massive dataset, the headline result was clear: AI-generated ads performed statistically equivalently to human-made ads in terms of CTR. Under loose controls, some AI creatives even showed slightly higher CTRs. Even when the researchers tightened controls to compare like-for-like closely, performance remained similar. Crucially, the study also found no evidence that AI-generated ads attracted low-quality, curiosity-driven clicks — a major concern among advertisers. Conversion rates remained stable regardless of creative origin.

This suggests that the efficiency and scale benefits of AI — producing many variants quickly — don’t necessarily come at the expense of engagement or effectiveness. For performance marketers stretched by tight budgets and rapid campaign cycles, that is a significant validation.

Perception Matters More Than Origin

Perhaps the most fascinating insight from the research is less about AI vs. human and more about how consumers perceive ads. The study included a perception experiment where users judged whether an ad was AI-generated or human-made. The results showed a clear pattern:

• Ads perceived as AI-generated actually underperformed — even when they were human-made.

• Conversely, AI ads that were perceived as human-made delivered the highest engagement of all groups, outperforming both clearly human and clearly AI-looking creatives.
  

This reveals a subtle but critical point: consumer reactions are driven more by visual cues and creative execution than by the actual creative origin. Whether an ad was made by an algorithm or a designer mattered less than what it looked and felt like to the audience.

What Makes Ads Feel “Human”?

The study dug into the creative elements that drove engagement, and one factor stood out: the presence of large, clear human faces. Human faces proved to be the strongest cue for perceived authenticity and trust, significantly boosting CTR. Interestingly, AI tools — at least in this dataset using Taboola’s GenAI Ad Maker — were more likely to include prominent human faces than human designers, which partly explains why AI ads did so well in certain contexts.

In contrast, ads that appeared overly artificial — with heavy colour saturation, overly polished visuals, strong symmetry or other “AI-looking” aesthetic traits — tended to be penalised by audiences. These cues reduced engagement regardless of whether the ad was AI-generated or human-created.

Creative Strategy Still Matters

Although the overall findings are promising for AI adoption, the study also showed that industry context matters. Some sectors, like personal finance and food and drink, showed stronger performance boosts with AI ads, while categories like education showed more muted results. This suggests AI isn’t a one-size-fits-all solution — marketers need to strategically deploy it where it aligns with audience expectations and category norms.

Implications for Marketers

For performance marketers, the takeaway is both pragmatic and forward-looking:

1. AI can match human creativity at scale. Advertisers can generate thousands of creative variants quickly without sacrificing engagement or conversions.

2. Perceived human-ness drives engagement. Ads that feel authentic and relatable outperform those that look artificial — regardless of who created them.

3. Creative execution still matters. The best results come from focusing on design principles audiences respond to, like human faces and emotional resonance, not just automation.

4. Use AI strategically. Different industries and audiences respond differently to AI creatives, underscoring the importance of audience research and testing.
   

Conclusion: A New Phase of AI in Advertising

This study marks a turning point in how the marketing world thinks about generative AI. For too long, debates about AI ads have been theoretical or anecdotal. Now, real-world evidence shows that AI can deliver performance on par with human creatives — provided marketers leverage it with smart strategy and careful design.

In an era where speed, scale and cost efficiency are mission-critical, AI-generated ads offer an exciting tool in the marketer’s toolkit. But the real art — and the real advantage — will come from blending data-driven automation with human-centred creative insight.