Can AI-Generated Ads Move Us? What EEG and Eye-Tracking Revealed About Viewers’ Responses

What happens when viewers are emotionally moved by an advertisement—only to learn afterward that it was created by AI?

Generative AI video is becoming increasingly sophisticated. Yet many people still believe that AI cannot reproduce the emotional authenticity rooted in human experience. This question is particularly important in advertising. An advertisement does more than communicate information. It is designed to help viewers empathize with its characters, become immersed in a story, and ultimately connect those feelings to a brand.

Can an AI-generated video advertisement capture attention and evoke emotion in the same way?

A study first published online in Psychology & Marketing in July 2026 examined this question by combining EEG, eye-tracking, an affect scale, and in-depth interviews.

The results cannot be reduced to a simple contest between humans and AI. The AI-generated advertisement elicited stronger emotional responses than the researchers had expected, and some participants who viewed it revised their perceptions of AI’s emotional capabilities in a more positive direction.

However, the study was an exploratory experiment involving only 20 university students. Its findings should not be generalized as evidence that AI-generated advertising can replace or outperform human-created advertising.

Why Did the Researchers Measure Emotion?

Previous evaluations of AI-generated advertising have often relied on surveys. The problem is that what people say about an advertisement does not always correspond with how they respond while watching it.

A viewer may claim that “AI-generated videos have no emotion” while still focusing intensely on a particular scene or exhibiting a strong physiological response. Conversely, viewers may believe that they were immersed in a story while failing to remember the brand or its services.

To examine this discrepancy, the researchers combined dual-process theory with the theory of algorithm aversion.

Dual-process theory distinguishes between an intuitive and emotional route of information processing and a more deliberate, analytical route. Algorithm aversion refers to people’s tendency to distrust or undervalue algorithmic outputs, particularly in areas believed to require creativity, empathy, or subjective judgment.

The researchers began with a clear expectation: AI-generated advertisements would stimulate more analytical than emotional processing, and knowing that an advertisement had been created by AI would interfere with emotional immersion.

Creating AI and Human Advertisements From the Same Story

The study included 20 international university students between the ages of 18 and 34. Nine participants were male and 11 were female. They were assigned either to an AI-generated advertisement group or a human-created advertisement group, with each participant viewing one version of the advertisement. The data presented in the article indicate that each group consisted of 10 participants.

To reduce the influence of prior brand familiarity or loyalty, the researchers created a fictional brand called “TruePath Career Services.” The advertisement told the story of four characters who initially conformed to social pressure before embracing their authentic identities and pursuing their passions.

The AI-generated and human-created versions used the same script, music, slogans, narrative structure, and overall composition.

The AI version was generated using Hailuo AI Video Generator. For the human-made version, university theatre performers recreated the scenes shown in the AI-generated video. Both versions were edited with VN Video Editor, while Audacity was used for audio recording. A consistent brand logo, created with Ideogram, appeared at the end of the advertisements.

By standardizing the content of the two videos, the researchers sought to focus on differences related to the production method.

One aspect of the design nevertheless requires attention. The human advertisement was not the original work that was subsequently reproduced by AI. Instead, human performers recreated scenes that had first appeared in the AI-generated version. This should be considered when interpreting the comparison.

The AI Origin Was Disclosed Only After Viewing

One of the most notable features of the experiment was the timing of the disclosure.

Participants did not know whether the advertisement they were watching had been generated by AI or produced by humans. The production method was disclosed later, during the post-viewing semi-structured interview.

This delayed-disclosure design was intended to reduce the immediate influence of an “AI-generated” label. It enabled the researchers to distinguish between participants’ initial reactions to the video and their subsequent evaluations after learning who—or what—had created it.

During the viewing session, EEG was used to record indicators associated with attention, engagement, excitement, interest, relaxation, and stress. Eye-tracking measured how long participants looked at predefined areas of interest, including the characters’ faces and actions central to the narrative.

Changes in affect before and after viewing were measured with the PANAS-SF scale. The subsequent interviews examined participants’ emotions, their interpretation of the advertisement, their attitudes toward AI, and their perceptions of the brand.

The qualitative interview data were coded using the Gioia methodology. An independent coder who was not part of the research team also participated in the coding process. Intercoder reliability, measured using Krippendorff’s alpha, was 0.761.

The AI Advertisement Evoked More Emotion Than Expected

The researchers’ first proposition predicted that AI-generated advertising would activate analytical thought more strongly than emotional processing. The findings did not support that expectation.

Participants who watched the AI advertisement understood its central message. At the same time, most spoke primarily about emotional experiences—including inspiration, understanding, sadness, and pressure—rather than offering only analytical evaluations.

In other words, the AI-generated advertisement stimulated cognitive processing, but it also evoked emotional responses more frequently than the researchers had anticipated.

The EEG data indicated generally high engagement within the AI-advertisement group. For some participants, engagement scores rose above 90 on the study’s measurement scale. Two participants who were familiar with AI technology showed particularly high levels of engagement.

These figures, however, do not establish that AI advertising was superior on average. The EEG findings were presented mainly as exploratory ranges for individual participants rather than as results from a large, confirmatory statistical comparison between populations.

Eye-tracking showed that most participants focused on the main characters and the key narrative elements. At the same time, nearly half of the participants in the AI group displayed more erratic gaze patterns. The researchers suggested that some visual elements or parts of the storyline may have been confusing or lacked a clear focal point.

No Statistically Significant Difference in Affect Was Found

The human-created advertisement also produced substantial emotional engagement. Participants responded to the real actors, music, and narrative, and some connected the story to their own lives or to broader social issues.

Their gaze was generally concentrated on the main characters and central narrative elements rather than on peripheral areas of the screen.

However, the study did not find statistical evidence that the human-created advertisement was clearly superior in terms of affect.

The difference between the two groups in positive affect was not statistically significant (p = 0.909). The difference in negative affect was also not statistically significant (p = 0.160).

This does not demonstrate that the two advertisements have identical emotional effects. With only 20 participants, the experiment had limited statistical power to detect differences between the groups.

A more accurate interpretation is therefore not that “AI and human advertisements have the same emotional impact,” but that this particular experiment did not confirm the assumption that AI-generated advertising would fail to evoke emotion.

Evaluations Shifted After the AI Origin Was Revealed

An intriguing change appeared after the source of the advertisement was disclosed.

Some participants who had watched the AI advertisement reported a sense of coldness or distance after learning how it had been created. The AI label may therefore function as a kind of “cognitive brake,” encouraging viewers to reassess an emotional experience through a more analytical lens.

Most participants, however, did not display strong algorithm aversion. Many were surprised that AI could produce such an emotionally engaging video. Some revised their opinions of AI’s capabilities in a more positive direction. Others perceived the brand as innovative because it had used AI to produce the advertisement.

By contrast, participants who had watched only the human-created advertisement expressed relatively consistent doubts that AI could reproduce the emotional depth and authenticity of human advertising. Although they acknowledged AI’s technical progress, they generally preferred human-created content and emphasized the genuine emotions and effort associated with human creators.

This contrast suggests that perceptions may differ between people who directly experience AI-generated creative work and those who evaluate its potential without seeing it.

Nevertheless, the two groups viewed different videos. The study did not use a within-subject design in which the same individual watched and compared both versions. The findings therefore cannot establish that direct exposure itself reduced algorithm aversion. They indicate only that such a possibility was observed and deserves further investigation.

A Moving Advertisement Is Not Necessarily an Effective Brand Advertisement

Perhaps the most significant finding concerns the relationship between the story and the brand, rather than the identity of the creator.

Both the AI-generated and human-created advertisements evoked emotion, but many participants had difficulty connecting the emotional story with the career-service brand. Some became so immersed in the narrative that they failed to notice the logo. Others assumed that the advertisement represented a different kind of company.

Emotional engagement did not automatically translate into brand trust or recall.

Participants whose own circumstances were relevant to the career-service message—such as those actively thinking about employment—showed greater interest in the brand. Those who believed the advertisement lacked information or that its emotional narrative was poorly aligned with the service were less willing to interact with the company in the future.

This finding draws attention to a frequently overlooked issue in discussions of AI video production. An advertisement does not achieve its purpose simply because it is visually impressive or emotionally stimulating.

The relationship between the story, the brand, and the product or service may be more important than whether the video was produced by AI or by humans.

How Far Can These Findings Be Trusted?

The study is notable for combining EEG, eye-tracking, an affect scale, and interviews. This multimethod approach enabled the researchers to examine immediate physiological responses and patterns of visual attention that would have been difficult to capture through survey responses alone.

Its limitations, however, are substantial.

The study included only 20 university students, with 10 participants in each condition. It is not known whether the same results would emerge among older adults, other demographic groups, or actual consumers in different market settings.

The laboratory environment may also have altered natural viewing behavior. Watching an advertisement while wearing measurement equipment is different from encountering it on a smartphone while scrolling through multiple pieces of content.

Although EEG and eye-tracking offered physiological and behavioral data, the study also relied on self-report measures such as PANAS-SF and open-ended interviews. These responses may have been influenced by social desirability, self-censorship, or participants’ beliefs about the creative legitimacy of AI.

The timing of disclosure represents another limitation. Participants were told about the AI origin only after watching the advertisement. In real advertising environments, AI involvement may be disclosed before viewing or presented alongside the content. The authors therefore call for future experiments that manipulate disclosure timing.

Most importantly, the researchers explicitly state that their findings do not yet identify the conditions under which AI-generated advertisements may or may not outperform human-created ones.

This is not a study proving the superiority of AI. It is an exploratory study that challenges the assumption that AI-generated video cannot evoke emotional responses.

What Matters Is Not Only Who Created It, but What Viewers Feel and Remember

This study does not determine whether humans or AI are more creative. Instead, it demonstrates that the way viewers process creative content is more complicated than a simple human-versus-machine comparison.

People may respond first to the music, characters, visual composition, and transitions in an AI-generated video. Only later, after learning its origin, might they reconsider its authenticity and production method. Emotional engagement may therefore emerge before information about AI authorship begins to shape the evaluation.

At the same time, a video that evokes emotion is not necessarily an effective advertisement. A powerful story may attract attention, but if it is disconnected from the brand, viewers may not remember what the advertisement was promoting or what action they were expected to take.

As AI-generated video becomes more common, the relationship between narrative purpose and brand meaning may become just as important as production technology.

The question is no longer only whether AI can create emotion. We also need to ask what that emotion was designed to accomplish and where it leads in the viewer’s memory.

Can AI-generated advertising move people?

This study offers a cautious answer: it may be possible. Whether that emotional response develops into trust, recall, or action, however, remains a matter that human creators must continue to design, examine, and verify.