AI-generated video advertising produced emotional reactions among viewers in a laboratory study of 20 university students, but the small experiment cannot establish whether AI ads outperform human-made campaigns in the wider market.
Researchers compared two versions of an advertisement for a fictional career service, one generated with artificial intelligence and the other performed by people. They used eye tracking, electroencephalography (EEG), questionnaires and interviews to examine viewers’ reactions.
The study in Psychology & Marketing, published online in July 2026, found emotional responses to the AI-produced video. Some participants also struggled to connect the story to the advertised service, a concern the authors observed for both versions.
What the researchers measured
EEG records electrical activity at the scalp. Eye tracking records where viewers look. Neither measure reads a person’s thoughts or proves that an advertisement will generate a sale.
The researchers, Mariia Shitova, Agnieszka Kabalska, Ákos Varga and Ralf Wagner, studied 20 international university students aged 18 to 34. The fictional brand and job-search theme were chosen for relevance to the participants.
According to the paper, the two versions shared a storyline, music and composition. Participants were told about AI authorship after viewing, so their initial reactions occurred before that disclosure.
The authors describe their work as exploratory. Two videos and a small student sample cannot represent the wide variety of audiences, products and creative choices involved in commercial advertising.
Attention is not the same as understanding
Viewers can remember a scene yet miss the name of the company or what it sells. In the experiment, some participants found a mismatch between the emotional story and the career service’s purpose.
That result gives marketers a concrete question to test: can the intended audience identify the brand and explain the offer after seeing the advertisement? An ad may attract attention while leaving viewers unsure what action to take.
The researchers reported a range of reactions to the AI version, including positive feelings and reservations. The authors’ explanation of those responses is not a measurement of future sales or proof that AI origin makes a brand more trusted.
Testing a campaign before scaling it
Businesses comparing ads can use matched audiences, similar budgets and consistent measurement periods. They can assess brand recognition and message comprehension as well as viewing time, clicks and completed purchases.
Costs matter too. Production savings from generative software should be weighed against editing, rights checks, review time and the cost of correcting misleading or unusable material.
Our earlier coverage of how previous campaign evidence can help evaluate AI marketing messages examined email campaigns, a different format and research question.
The new video study offers an early look at audience reactions, not a general verdict on human or AI creativity. A larger test across campaigns and customer groups would be needed to establish how reliably the findings carry over.
