Scroll through Instagram or TikTok and you’ll increasingly encounter influencers who don’t exist. Computer-generated personalities like Lil Miquela have racked up millions of followers, and global brands including BMW, IKEA, and Louis Vuitton have hired virtual endorsers to pitch products. But can a digital avatar actually make you feel something the way a human spokesperson can?
A new investigation published in Psychology & Marketing suggests the answer depends heavily on the kind of feeling being sold. Virtual endorsers can convey vivid, in-the-moment emotions like excitement or joy about as well as flesh-and-blood humans. But when the pitch calls for something more reflective, like gratitude, hope, or pride, the digital stand-ins fall noticeably short.
The question behind the research
Prior research on virtual endorsers has produced conflicting results. Some studies concluded that audiences see digital influencers as lacking genuine emotional experience, which makes them poor fits for emotionally charged campaigns. Other studies found that lifelike gestures and facial expressions from virtual endorsers can still move audiences. The findings didn’t line up.
Chenying Hai and colleagues at the School of Management at Huazhong University of Science and Technology in Wuhan, working with a co-author at John Carroll University in Ohio, suspected the mixed results came from treating positive emotions as a single category. Feeling excited about opening a birthday gift and feeling grateful for a lifelong friendship are both positive, but they operate quite differently in the mind.
The researchers drew on a framework that splits emotions into two types. Concrete emotions, like excitement and happiness, are tied to specific events and produce vivid, immediate, bodily reactions. Abstract emotions, like gratitude, hope, or a sense of fulfillment, come from broader interpretations of life circumstances and produce quieter, more contemplative responses. The team wanted to know whether virtual endorsers might handle one kind better than the other.
Mining social media, then running experiments
The investigation began in the wild. The researchers scraped 10,704 posts from Rednote, a Chinese lifestyle-focused social media platform with hundreds of millions of monthly users. They collected content from the top 10 virtual endorsers on a widely cited industry ranking and, for comparison, from 50 human endorsers on the same platform.
Because no existing dictionary reliably distinguishes concrete from abstract emotions, the team used OpenAI’s GPT-5 through its programming interface to classify each post’s emotional content. They built a structured prompt that included definitions, examples, and step-by-step decision rules, and then compared the model’s outputs against human coders to check reliability.
With posts sorted, the researchers used the number of “likes” as a measure of how audiences responded, applying a regression model that controlled for factors like follower counts, hashtags, emojis, post length, gender of the endorser, and time of posting.
The pattern in the data was clear. When posts conveyed concrete emotions, virtual and human endorsers pulled roughly comparable engagement. But when posts leaned on abstract emotions, virtual endorsers received significantly fewer likes than their human counterparts. The interaction held up across several robustness checks, including alternative measures of engagement and different statistical models.
Testing the pattern in controlled experiments
To move beyond correlation, the team ran five controlled experiments involving more than 1,700 participants recruited in China and the United States. Each experiment showed participants an advertisement featuring either a human or virtual endorser, pitching products that ranged from headphones and cameras to a fitness app and a Disney trip. The emotional appeal varied between concrete (excitement, joy) and abstract (gratitude, hope, fulfillment).
The results mirrored the field data. When advertisements used concrete emotional appeals, participants rated ads featuring virtual endorsers just as favorably as those featuring humans. When ads used abstract appeals, the virtual endorsers consistently underperformed. This held true regardless of whether the virtual endorser looked highly realistic or more stylized and cartoonish, whether the ad was a static image or a video, and whether the endorser was male or female.
Two pathways behind the gap
The researchers wanted to understand what was driving the difference. Drawing on a theory that emotional expressions communicate information through two parallel routes, an emotional one and a cognitive one, they measured both.
On the emotional side, they looked at emotional contagion, the tendency for people to unconsciously catch and share the feelings someone else expresses. On the cognitive side, they measured expectation violation, or the sense that something feels off or inappropriate about what an endorser is expressing.
For concrete emotions, both endorser types produced similar levels of emotional contagion and similarly low levels of expectation violation. For abstract emotions, virtual endorsers produced weaker emotional contagion and stronger expectation violation. Participants seemed to find it harder to catch abstract feelings from a digital face, and were more likely to sense something was amiss about the display.
The researchers interpret this as reflecting the nature of abstract emotions themselves. These feelings depend less on immediate, observable cues like a smile or a jump, and more on inferred meaning and authenticity. Because virtual endorsers lack lived experience, audiences appear to be more skeptical when they claim to feel something reflective or values-based.
A workaround: shared attitudes
A final experiment tested whether anything could close the gap. The researchers manipulated how similar participants felt to the endorser in terms of attitudes and values. In one condition, participants learned the endorser was an environmentalist, and then rated how important environmental protection was to them personally. Those who cared strongly about the environment saw the endorser as sharing their values.
When perceived attitude similarity was high, the virtual endorser’s disadvantage in conveying abstract emotions largely disappeared. Emotional contagion strengthened and expectation violation weakened. When similarity was low, the gap remained wide.
Notably, the researchers found that surface-level similarity, like how human-like the virtual endorser looked, did not have the same compensating effect in prior related research. Making a digital avatar look more realistic didn’t help it convey abstract emotions any better. Sharing values and beliefs mattered; sharing pixel-level realism did not.
What this means for marketers
The findings offer some direct guidance for brands considering virtual endorsers. Campaigns built around vivid, immediate emotional payoffs, the thrill of a new gadget, the fun of a product experience, the delight of a gift, appear to work about equally well with virtual or human faces. Campaigns built around deeper, values-based emotions, gratitude for community, hope for the future, pride in personal growth, still favor human endorsers.
If a brand does want to use a virtual endorser for an abstract emotional pitch, the research suggests one workaround: engineer the endorser’s persona to align with the target audience’s values and beliefs. Positioning a virtual influencer as an environmentalist, a fitness enthusiast, or an advocate for a specific cause appears to help audiences bridge the authenticity gap.
The authors note several limitations. The field data came from a single Chinese platform, and the experiments focused on positive emotions rather than the full emotional spectrum. The study also didn’t account for the parasocial relationships that develop between audiences and popular virtual influencers over time, which could change how their emotional expressions land. And large language models, while efficient for classifying emotional content at scale, aren’t yet perfectly reliable coders of subtle emotional distinctions.
