The past few years have been a living experiment on broadly adopting a new technology — generative AI — for businesses and individuals. It’s largely been a blank canvas, with people exploring how to use AI to create written drafts, videos, imagery, and more, alongside process automation, data exploration, and other practical applications.
AI usage in advertising only continues to rise, with 83% of ad execs saying that their company has deployed AI in the creative process (that’s up from 60% in the same source’s 2024 study).
Now, we’re finally capturing data to truly understand how gen AI compares to solely human-created work. In the world of digital advertising, a new study from Columbia, Harvard, Technical University of Munich (TUM), and Carnegie Mellon, found that AI-generated ads perform at the same level as, or better than, human-made content. The researchers analyzed more than 300,000 ads and more than 500 million impressions. The AI-generated ads studied were created using Realize’s GenAI Ad Maker, which lets advertisers create ads using AI alongside their traditional, human-made content.
The study found some key parameters to help advertisers understand how to use AI and human-made work in tandem, and how to incorporate the benefits of AI to serve up great ads. Use these best practices for creating and testing AI-generated ads that connect with your audience.
Columbia Study Form
1. Prioritize Clear Human Faces in the Frame
When it comes to AI-generated ads, the single most powerful visual feature for driving engagement is the presence of faces. Consumers recognize this signal of humanness, and they strongly associate larger displayed faces with human-made content. That perception increases trust and drives a higher click-through rate (CTR).
Design elements matter quite a lot for performance when creating ads, per AdExpert AI: Genuine smiles increase engagement by 27% compared to neutral expressions, and authentic, candid shots outperform professional studio photos by 41%.
Newer AI tools are already incorporating this parameter into their creative work. Realize’s GenAI Ad Maker, for example, is actually more likely to include human faces than traditional human-made ads, aligning with long-standing creative best practices. When prompting AI on imagery, specify “close-up” or “mid-shot” views to ensure facial features are prominent and clear.
2. Master the Looks-Like-AI Spectrum
Our human brains often gauge imagery as AI-generated based on a quick glimpse, leading to the “uncanny valley” effect. That’s when there’s a drop in human empathy and comfort upon seeing artificial beings that are almost human, but not quite. It can apply to robots, CGI, or even dolls, and it’s important to consider in AI-generated ads.
To make sure imagery is comfortable to consumers, try these tips:
Avoid over-saturation
Consumers often identify ads as AI-generated if they feature intense color saturation or overly polished aesthetics.
Aim for realism
Higher levels of perceived artificiality significantly lower an ad’s performance. AI ads that were not perceived as AI achieved the highest CTR of all groups in the Columbia study, scoring even above those that were actually made by people.
Disguise through quality
Use features that consumers mistakenly associate with humans, such as high image clearness and sharpness, to “disguise” the AI’s origin.
Ultimately, AI-generated ads perform best when they feel authentically human.
3. Consider Strategic Industry Application
Advertisers and performance marketers work in incredibly varied product categories, and they’re not all created equal. The effectiveness of AI varies significantly by product category, so advertisers should use AI where it shows the strongest performance.
The research study found that AI-generated imagery has delivered especially strong results in personal finance, pets, real estate, and food and drink, so try AI-created ads in these high-performing verticals. Conversely, categories like education haven’t seen the same uplift, suggesting that audience expectations for human expertise are higher in these fields.
Recent data has also found that retail and e-commerce is the top industry for use of AI in advertising, with 26% using AI, more than financial services, media, healthcare, and others. Consider AI ads with contextual sensitivity based on what you know about your audience.
4. Use the Sibling Ads Testing Framework
In the study, the researchers used a “sibling ads” approach, comparing matched pairs of AI-generated and human-made ads created by the same advertiser, for the same campaign, on the same day, with the same objective and landing page. This approach let them isolate the impact of AI-generated visuals while controlling for all the many other variables that typically influence ad performance.
For your own ad testing and to maximize ROI, move beyond simple A/B testing and try the sibling ads approach for faster iteration. Launch AI-generated and human-made variations simultaneously within the same campaign, sharing the same landing page and objectives. Because AI lets you create dozens of ad variations in minutes, you can identify the winners faster, without the creative burnout of traditional design cycles.
5. Drive Clicks Without Sacrificing Conversions
One recent concern for performance marketers is that AI ads might only drive curiosity clicks that don’t lead to true engagement. But, in reality, the data shows that AI-generated visuals increase or maintain CTR without reducing downstream conversion performance.
There’s a lot of potential for AI to improve ad performance through engagement, too. Microsoft found that users of Copilot engage much more with ads, driving 73% higher CTR and 16% stronger conversion rates. Plus, Copilot shortens customer journeys by 33%, reducing the number of steps to conversion versus traditional search.
It’s also important to remember that, because AI ads can be produced at zero or near-zero cost, the on-par performance results in a significantly higher overall ROI for the campaign.
Key Takeaways
As AI usage continues to grow in creative work like advertising, there’s more and more data available on how AI-driven ads perform compared to human-created ads. The future of high-performance advertising will rely on a strategic blend of AI efficiency and human-centric design. Advertisers can continue improving ad effectiveness and cutting down on production overhead by prioritizing authentic human cues, avoiding perceptual markers, and using elements that consumers are comfortable with. As models continue to improve, the most successful brands will be those that use AI to enhance, not replace, the human feel of their creative work.
Frequently Asked Questions (FAQs)
Does using AI-generated imagery negatively impact my ad’s click-through rate (CTR)?
There have been concerns among advertisers that algorithm aversion will cause users to ignore or distrust AI content. But, within digital advertising, performance depends heavily on the quality of the output and how well it mimics human-made design principles. The Columbia study, which studied Realize’s AI ad creator specifically, showed that AI-generated ads actually achieved a higher average CTR (0.76%) compared to human-made ads (0.65%) in the full dataset. Using the sibling ad method with identical settings, AI ads performed the same or better against human-made ones, proving they do not sacrifice performance for efficiency.
Is there a risk that AI ads will attract junk clicks that don’t lead to actual sales?
Another recent industry concern is that AI content might drive curiosity clicks, where users click on an ad simply because the image looks strange or novel, leading to high bounce rates and low conversions. In the research study, however, the data found no evidence of this curiosity click phenomenon. AI-generated visuals on the platform increased or maintained CTR without reducing downstream conversion rates, meaning advertisers gained scale without trading off quality or ROI.
What is the most effective way to ensure an AI-generated ad performs well?
Successful AI ads should continue to follow established marketing psychology. That includes using clear focal points and high-quality resolution, while avoiding the uncanny valley effect where images look nearly, but not quite real to human viewers. AI ads perform best when they do not look like AI to the consumer. Those using Realize for AI ad creation know that the art of disguise is key to success: Make sure that ads built with AI include large, clear human faces, which is a feature that consumers strongly associate with human-made content, and that drives significantly higher engagement.
