After reading, you will be able to:

• define a virtual influencer and a virtual idol,
• explain para-social interaction and the uncanny valley,
• describe how facial realism, attractiveness and familiarity change brand attitude,
• read basic results from a PLS-SEM study and an experiment.

What are virtual influencers?

A virtual influencer (VI) is a digital character created with computer graphics (CGI) or generative AI. It has a name, a personality and a social media profile. A team of designers, writers and marketers stands behind it. The character posts content, talks to followers and signs brand deals.

In Asia, a closely related term is virtual idol. Yu, Bannasilp and Kwong (2025) define a virtual idol by three elements:

1. Digital entity – it exists only as data, not as a living person.
2. Anthropomorphic appearance – it looks human-like (anthropomorphism means giving human traits to non-human things).
3. Idol activity engagement – it performs like a celebrity: sings, streams, gives concerts, endorses brands.

Virtual idols started in Japan in the 1990s. The best-known one is Hatsune Miku, a singing voice software character from Crypton Future Media (2007) who now fills concert halls with hologram shows.

Types of virtual influencers

Real business cases of virtual influencers

• Lu do Magalu (Magazine Luiza, Brazil). Lu was created in 2003 as a shopping assistant for the retailer’s website. Today she is a full brand personality. In 2022 the ranking site Virtual Humans named her the most-followed virtual influencer in the world, with about 55 million followers across platforms (Forbes Brasil, 2022). This is a classic case of a company owning its influencer.
• Lil Miquela (USA). Launched in 2016 by the start-up Brud, Miquela has worked with fashion brands such as Prada and Calvin Klein and was signed by the talent agency CAA (Wikipedia).
• Aitana López (Spain). Created by the Barcelona agency The Clueless, she gained over 340,000 Instagram followers in about 18 months. Her creators say she earns around €3,000 a month, and up to €10,000 in good months (Euronews, 2024).
• Luo Tianyi (China). A virtual singer with 5.4 million Weibo followers. Her endorsements cover food, telecoms, insurance, banking and cars (Yu et al., 2025).

Why do brands like them? Human celebrities are expensive and can cause scandals. A virtual influencer never gets tired, never ages and never posts something embarrassing at 3 a.m. – the brand controls every word (Moustakas et al., 2020, cited in Xian et al., 2026). The market is large: iiMedia Research estimates that industries driven by virtual humans in China will reach 640 billion RMB in 2025 (cited in both papers).

Study 1: Virtual idols and para-social interaction

The first paper is “Virtual Idols’ influence on consumer’s brand attitude and purchase intention: A perspective of para-social interaction” by Yu, Bannasilp and Kwong, published in Electronic Commerce Research and Applications (2025).

Key term: para-social interaction (PSI)

Para-social interaction is a one-sided relationship. A fan feels that they know a media figure personally, like a friend, but the figure does not know the fan. You might feel this way about a YouTuber you watch every day. The study asks: can people build such a bond with a character who is not even human?

Method

The authors surveyed 418 fans of virtual idols on Chinese platforms (Weibo, Bilibili, Baidu Post Bar). 92.6% were Generation Z (aged 12–27). They used PLS-SEM (Partial Least Squares Structural Equation Modelling) – a statistical method that tests many cause-and-effect paths between hidden concepts (like “attractiveness”) measured by survey questions.

Results

Four features of a virtual idol build para-social interaction. The number β (beta) shows the strength of each path: higher means stronger.

Homophily means “love of the same” – we like people who are similar to us. The lesson is surprising: values matter more than looks.

The second big finding is about mediation. A mediator is a variable that carries an effect from A to C. Here, PSI affects purchase intention mostly through brand attitude. Fans know a virtual idol does not really use the product. So the idol does not convince them to buy directly; instead, their affection transfers to the brand (affect transfer), and a better brand attitude leads to buying. The authors’ advice: make brand attitude, not immediate sales, the main goal of a virtual idol campaign.

Python: decomposing the total effect

The indirect effect equals the product of two paths (a × b). The paper reports total = 0.467 and indirect = 0.317. Let us check what share of the effect goes through brand attitude.

# Yu et al. (2025) - mediation decomposition
a = 0.457          # PSI -> brand attitude
indirect = 0.317   # PSI -> brand attitude -> purchase intention
direct = 0.150     # PSI -> purchase intention

b = indirect / a   # implied path brand attitude -> purchase intention
total = direct + indirect
share = indirect / total

print(f"Implied b path: {b:.3f}")          # 0.694
print(f"Total effect:   {total:.3f}")      # 0.467
print(f"Mediated share: {share:.0%}")      # 68%

About two-thirds of the influence goes through brand attitude. This number is called the variance accounted for (VAF), and it is a common way to describe partial mediation.

Study 2: Virtual influencers, facial realism and the uncanny valley

The second paper is “Beyond the uncanny valley: the dual pathways of virtual influencers’ facial realism on brand attitude” by Xian, Zhang, Liu, Huang and Zhang, also in Electronic Commerce Research and Applications (2026). It asks a design question: how human should a virtual influencer’s face look?

Key terms

• Uncanny valley – a theory by robotics professor Masahiro Mori (1970). As a robot or avatar looks more human, we like it more, until it looks almost human. Then small errors (dead eyes, plastic skin) make it feel creepy. Liking drops into a “valley”.
• Eeriness – the creepy, uncomfortable feeling the uncanny valley produces.
• Meaning transfer theory – Grant McCracken’s (1989) idea that the cultural meanings of an endorser (e.g. “cool”, “luxurious”) move to the brand they endorse.
• Coolness – being seen as unique, trendy and attractive in a slightly rebellious way.

Method: four experiments

The authors ran four online experiments with Chinese consumers (N = 298, 289, 253 and 247). Participants saw an ad for a fictitious tech product (phone, headphones, laptop) endorsed by a virtual influencer with a high- or low-realism face. In an experiment, people are randomly assigned to groups, so differences between groups can be read as causal effects – something a survey cannot prove.

Results: two pathways

1. Main effect. A realistic face improved brand attitude a lot: 5.07 vs. 3.87 on a 7-point scale (Study 1).
2. Positive “coolness” path. A realistic VI looks cooler → the brand looks cooler → brand attitude rises. This is meaning transfer in action.
3. Negative “eeriness” path. A low-realism face felt eerie, and eeriness lowered brand attitude directly. It did not work through brand authenticity, as the authors expected. Eeriness is a fast gut reaction, not a rational judgement.

Boundary conditions: attractiveness and familiarity

A boundary condition (moderator) changes the strength of an effect.

Two lessons stand out. First, the halo effect – our tendency to assume that beautiful people are also good in other ways – partly protects an imperfect face. But beauty did not remove eeriness. A pretty but “wrong” face can still feel creepy. Second, the familiarity result was the opposite of the hypothesis. If a VI’s face is modelled on a famous person, consumers study it carefully. Perfect execution is rewarded; weak execution is punished.

Python: effect size (Cohen’s d)

Cohen’s d tells us how big a difference between two groups is, in standard deviations. Rules of thumb: 0.2 small, 0.5 medium, 0.8 large.

import math

# Xian et al. (2026), Study 1: brand attitude
m_high, sd_high = 5.07, 1.19   # high facial realism
m_low,  sd_low  = 3.87, 1.46   # low facial realism

pooled_sd = math.sqrt((sd_high**2 + sd_low**2) / 2)
d = (m_high - m_low) / pooled_sd
print(f"Cohen's d = {d:.2f}")   # about 0.90 -> large effect

Our result (0.90) matches the paper’s reported d = 0.901. Facial realism is not a cosmetic detail; it has a large effect.

What do the two studies mean for marketers?

• Choose values before looks. Homophily was the strongest driver of fan bonds. Give the character a clear personality and values that match your target group.
• Keep it active. Regular posts, live streams and replies to fans build para-social interaction.
• Measure brand attitude, not only clicks. Virtual influencers work mainly through attitude. Use brand-lift surveys, sentiment analysis and long-term KPIs, not just last-click sales.
• Go fully realistic or clearly stylised. The danger zone is the middle. Invest in skin texture and eye details, or choose an openly cartoon style like Hatsune Miku.
• Be careful with look-alikes. A VI based on a famous face raises expectations.
• Be transparent. Label AI-generated content and sponsored posts. The EU AI Act and consumer-protection rules increasingly require it.

Limitations to remember

Both studies used Chinese samples, so culture may shape the results. Study 1 used self-reported survey data from fans who already liked virtual idols. Study 2 used fictitious tech brands only. Results may differ for food, fashion or utilitarian products.

Essay topics for students

1. Virtual influencers in Polish e-commerce: Could a brand like Allegro, InPost or Reserved build its own “Lu do Magalu”? Design the persona and a KPI plan.
2. Ethics of virtual influencers: Should AI-generated influencers be labelled? Discuss transparency, body-image pressure and the EU AI Act.

Related reading on this blog: AI influencers · Authenticity in Influencer Marketing · TikTok Shop and customer experience · Customer Journey in the Age of AI

Summary: three key questions

1. What are virtual influencers and why do brands use them?
They are computer-generated characters that act as social media creators and brand endorsers. Brands use them because they are fully controllable, scandal-free, always available and attractive to Generation Z.

2. How do virtual influencers influence consumers?
Through para-social interaction: fans form a one-sided emotional bond, driven mostly by perceived similarity in values (homophily), social attractiveness and activity. This bond improves brand attitude strongly (β = 0.457), and brand attitude then drives purchase intention (Yu et al., 2025).

3. Why does facial realism matter?
A realistic face transfers “coolness” to the brand, while an almost-but-not-quite human face triggers eeriness (the uncanny valley) that directly lowers brand attitude. High attractiveness softens this effect; a familiar face makes consumers more critical (Xian et al., 2026).

Sources

• Yu, Y., Bannasilp, A., & Kwong, S. C. M. (2025). Virtual Idols’ influence on consumer’s brand attitude and purchase intention: A perspective of para-social interaction. Electronic Commerce Research and Applications, 74, 101546. https://doi.org/10.1016/j.elerap.2025.101546
• Xian, X., Zhang, Y., Liu, R., Huang, C., & Zhang, L. (2026). Beyond the uncanny valley: the dual pathways of virtual influencers’ facial realism on brand attitude. Electronic Commerce Research and Applications, 79, 101628. https://doi.org/10.1016/j.elerap.2026.101628
• Forbes Brasil (2022). Por que a Lu do Magalu tornou-se a maior influenciadora virtual do mundo.
• Euronews (2024). Meet the first Spanish AI model earning up to €10,000 per month.
• Wikipedia. Miquela.
• McCracken, G. (1989). Who is the celebrity endorser? Cultural foundations of the endorsement process. Journal of Consumer Research, 16(3), 310–321.
• Mori, M., MacDorman, K. F., & Kageki, N. (2012). The uncanny valley. IEEE Robotics & Automation Magazine, 19(2), 98–100.

The post Virtual Idols: Why Digital Faces Sell Real Products appeared first on Customer Experience Management.

Based on source: Top Websites Ranking – Most Visited Websites in November 2023 | Similarweb

There are two ways to grow your company. One is to auire customers and another is to engage them. The only combination of both give you a balance in profitable growth. We can call it the growth through customer experience management.

About The Top Visited Websites

Question: Which is the most visited websites in the world?
Answer: google.com, youtube.com, facebook.com, instagram.com, twitter.com (the FIVE) are the most visited websites in November 2023. The first e-commerce website in 11th place is Amazon.com. Websites wikipedia.org, yohoo.com and yandex.ru are in 7-9 places.

I case of e-Commerce rankings in pure sellers we have gotten Amazon.com on the 11-th place and ebay.com on 49 as the eCommerce and Shopping > Marketplace

A interesting thing with Openai.com, which is on 19-th place, it is growing two positions up from previous month.

Question: Which website’s category is the most visited?
Answer: 

e-Commerce manager dashboard – lets take e-commerce pepper.pl

What they are – pepper is the place with the hottest deals. You can rate and comment them with our community. It presents deals, coupons, discount codes, price errors for what you are looking for like smartphones, laptops, consoles, toys, fashion, home and garden, groceries, etc. You can buy at the lowest prices and save. You can leverage community shopping on pepper.pl

Statistics

The most important statistics for e-Commerce manager dashboard.

Global rank, which is the traffic rank of site, compared to all other sites in the world is 2,630 and it grows 688 positions from previous month

Country rank, which is the traffic rank of this site shown for the country where it gets the most traffic, and it is for Poland 49 with the growth +16 m/m.

Category rank, which is the traffic rank of the site compared to all other sites in its main category in the top country. The category for pepper.pl is eCommerce & Shopping > eCommerce and Shopping – Other (In Poland), and the rank is 1.

Total visits, which is the sum of all visits on desktop and mobile from the last month, and it is 23 M. If we take amazon.com as ecommerce website for comparison, they have got 2.6B visits per month. Google search engine took 83.8B visits, and the next in ranking, youtube.com have got 31.4B visits.

Traffic Sources which is the top traffic source to pepper.pl. It is:

• Direct traffic driving 87.56% of desktop visits last month
• Referrals 0.23%
• Organic Search is the 2nd with 8.03% of traffic
• Paid Search < 0.01%
• Social 2,34%
• Mail 1.56%
• Display 0,27%

You can drill it down into the main traffic drivers in each channel.

You can also compare (benchmark) each channel of traffic acquisition for competitors. For example, the top competitor for pepper.pl is allegrolokalnie.pl. They have go 23.54% of visits from organic search which is 3 time more then pepper.pl got.

Bounce rate is the average percentage of visitors who view only one page before leaving the website. For pepper.pl it is 31,41% which means that every 3-th user goes out visiting just only one page of the website.

Pages per visit, which is the average page count per visit is 6.03. It looks pretty good.

Avg. visit duration, which is average duration of time spent on the site. For pepper.pl it is 8 min.

Engagement Analysis

Engagement Analysis show how well pepper.pl meets visitor expectations and captures their interest.

Top countries, which is the top countries sending traffic to the analyzed website on desktop. The majority of traffic comes from Poland, so it is a polish dedicated website.

Website Traffic Demographics, which is the audience divided according to gender ( female, male ) and age distribution. In case of pepper.pl, the audience is 70.13% male and 29.87% female and the largest age groups of visitors are 25 – 34 year old with share of 32% and 18-24 with the share of 25.5%.

Other Visited Websites – what other websites they visit: allegro.pl, wykop.pl, ceneo.pl

Top Categories, which is the distribution of the categories most visited by the analyzed website’s users. For pepper.pl you can see them on the picture below. It is according to similarweb.com for November 2023.

The post e-Commerce manager’s dashboard appeared first on Customer Experience Management.