Generative text-to-video AI (GenAI) models (e.g., Sora), capable of creating realistic and imaginative scenes from textual instructions, have caused a tectonic shift in the advertising industry. Brands and agencies are innovating at a breakneck pace to leverage AI-generated video content in their advertising. Behind the scenes, brands, agencies and researchers are evaluating the feasibility and risks of using AI-generated creative in the way an ad is imagined, produced and evaluated.

The time and money savings that AI-generated creative can bring during the ad campaign development process is very appealing to brands – why spend time and money creating an animation or storyboard when you can input text and watch the draft idea come to life in a matter of moments? Some even believe that AI-generated content will be able to eliminate the need to produce finished text and become the norm in the ad creation process. But are we there yet?

how does the brain process AI-generated advertising?

At first glance, some of the most powerful and highest quality AI videos may seem “real” At least, that’s what we think, or what we say we think...but what is the brain really telling us? Veteran neuroscientists at NIQ BASES Advertising sought to delve deeper into how the brain responds to AI-generated content, centering a study around two key questions:

We selected a set of AI-generated brand ads that fell on a spectrum from high to low quality. We then evaluated them using System 1 methodologies-electroencephalogram(EEG), eye tracking, implicit response time-in addition to traditional survey measures. In a forthcoming article, we will share our findings in more detail, but our initial findings raise important questions for marketing managers and advertisers about the use cases for generative AI.

AI-generated ads are often easily identifiable

Consumers were very attentive to the quality of the creative. When asked for their overall impressions of the ad, they identified most of them – spontaneously and without prompting – as AI-generated. In fact, the only ad that consumers did not immediately perceive as such was created by an advertising professional through considerable iterative editing. Consumers also rated all AI-generated ads as significantly more annoying, boring and confusing than non-AI-generated ads.

What this means for brands: AI-generated content can create a negative halo in consumers’ minds, even when they don’t explicitly perceive it as such. When using AI, brands should ensure that they are asked to exercise caution to deliver a quality end product.

AI-generated ads elicit weak memory activation

Compared to traditional video ads, we found that memory-measured in the brain with EEG-was weak in most of the AI-generated ads tested, even those perceived as “high quality”

Memory helps us understand whether what we are seeing fits with something we already know or have experienced, i.e., whether we have a template in our brain against which we can compare what we are seeing. This lower memory engagement was evident, even for the most edited and polished ad: Consumers ‘ brains detected that something didn’t fit, even if they weren’t aware of it.

AI-generated ads are good at triggering brand associations

Even lower-quality AI-generated ads were able to successfully convey the desired brand identity. Each of these ads was able to reinforce their respective brands’ mental network of associations at a strong level (measured by EEG in a pre-post exercise).

When generating images, the AI models draw on all existing representations of the brand from the training set (e.g., how the brand is presented in the media, online, etc.), essentially leveraging a visual stereotype of the brand. This makes brand associations immediately accessible, because our brains respond well to consistency and repetition. Often, creatives are looking to refresh their brand and move into new territory, which can take time and repetition to generate a strong brand association. AI, however, leverages the known and familiar.

AI-generated ads must be of high quality to effectively convey a message

Ad segments whose visual elements looked especially strange or unrealistic elicited the least memory activation and the highest attentional activation, as measured by EEG. This indicates that it takes a great deal of cognitive effort to understand what is happening in those moments.

The brain has a limited capacity to process information. If it is overloaded with complex or unrealistic images, key information can be lost. In addition, humans are incredibly good at detecting any deviations in a person’s stereotypical appearance and movements Uncanny Valley effect. This means that we may feel uneasy about AI and human-like robots that are close to but do not quite look like humans. We may experience a sense of strangeness, unease, and even fear in the presence of these objects.

what is the future of GenAI in advertising testing?

Our results indicate that consumers are very sensitive to the authenticity of advertisements, both at the implicit (non-conscious) and explicit (conscious) levels. Currently, even the best AI-generated videos fail to trick the brain, which reduces the effectiveness of ads. While this technology is not yet ready for prime time, it can help brand managers identify key brand assets, create story boards and generate insights at the earliest stages, which can greatly facilitate the ad development process. And as AI models become more and more sophisticated, their ability to produce more realistic video will likely only increase, giving agencies and marketers a valuable resource for creating new and effective ads.

Authors