Journal of Retailing and Consumer Services 71 (2023) 103233

Contents lists available at ScienceDirect

Journal of Retailing and Consumer Services

journal homepage: www.elsevier.com/locate/jretconser

Effects of human versus virtual human influencers on the appearance

anxiety of social media users
Fengyi Deng *, Xia Jiang
Chongqing Technology and Business University, Chongqing, China

A R T I C L E I N F O A B S T R A C T

Keywords: The association between social media use and an individual’s appearance dissatisfaction has generated broad
Social media influencer scholarly interests. This study explored the specific appearance dissatisfaction resulting from the effects of social
Virtual influencer media influencer images on the appearance anxiety of users, in particular the differences between human
Appearance anxiety
influencers (HIs) and emerging virtual influencers (VIs). Women (n = 178) aged 18–35 were randomly assigned
State appearance comparison
to HI images, VI images, and scenic images (control group) and were asked to answer related questionnaires. The
results showed that images of both HI and VI induced significantly higher appearance anxiety than in the control
group. The participants who were exposed to VI images reported significantly lower appearance anxiety than
those who were exposed to HI images. Participants’ state appearance comparison played a mediating role be­
tween the images of social media influencers and appearance anxiety. In addition, participants who viewed VI
images reported significantly lower state appearance comparison than those who viewed HI images. Insights for
influencer marketing practice were further discussed based on the results.

1. Introduction system which not only enables individuals to judge others’ and their
own appearance, but also serves to reinforce appearance ideals where
With the rapid development of social media, the dark side of influ­ they indicate social approval of those ideals. (Valle et al., 2021; Hawes
encer marketing has recently come under the academic spotlight (e.g. et al., 2020).
Seek and Barker, 2022; Naeem, 2021). Evidence from a meta-analysis of Previous research attention has shifted from discussing the effects of
24 experimental studies revealed that images of the ideal appearance social media use (e.g., Fardouly and Vartanian, 2016) to investigating
depicted on social media have a moderate negative effect on body image those of social media content (e.g., Slater et al., 2019; Rodgers et al.,
(Hedges’ g = − 0.61, p < 0.01; Valle et al., 2021). Appearance anxiety, 2021). However, the role of societal beauty standards has been over­
which refers to a function of responses (e,g, anxiety, tension, or looked. The visual presentations of influencers, friends, and strangers, as
nervousness) to body image, and shame, which includes various well as the comments they receive, cultivate social stereotypes and so­
cognitive processes, are among the important indicators of body images cietal beauty standards (Saunders and Eaton, 2018; Valle et al., 2021).
which have gained interest among many scholars (Cihan and Cihan, The societal beauty standards conveyed through social media are more
2019; Velkoff and Smith, 2020). Women who reported appearance influential than those delivered through traditional media, as people
anxiety are more likely to see themselves as objects (state tend to believe that social media reflect the reality surrounding them
self-objectification) to be viewed and evaluated based on their appear­ (Fardouly and Vartanian, 2016).
ance, causing a series of mental health risks (e.g., sensitive feelings, In the above mechanisms, social media influencers (SMIs) play an
negative mood, depression) (Dimas et al., 2021; Velkoff and Smith, 2020 important role as key opinion leaders regarding societal standards (Ge
). Many scholars have explore the role of social media in eliciting peo­ and Gretzel, 2018). The term SMIs refers to people who have “gained
ple’s appearance anxiety and have argued that social media may have a popularity due to their social media presence and content” (Wahab
significantly greater impact on appearance-related concerns than et al., 2022; Aw and Chuah, 2021). Although many studies have iden­
traditional media (Hawes et al., 2020). The user-generated content tified the association between social media and appearance anxiety,
(UGC) and real-time feedback features on social media form a complex none has highlighted the role of SMIs directly. The typical experimental

* Corresponding author.
E-mail addresses: [email protected] (F. Deng), [email protected] (X. Jiang).

https://doi.org/10.1016/j.jretconser.2022.103233
Received 17 August 2022; Received in revised form 10 December 2022; Accepted 12 December 2022
Available online 26 December 2022
0969-6989/© 2022 Elsevier Ltd. All rights reserved.
F. Deng and X. Jiang Journal of Retailing and Consumer Services 71 (2023) 103233

operation in existing studies involves asking participants to view ideal research results that ideal body pictures can elicit appearance anxiety
and non-ideal appearance images from social media (e.g., Carter and (Slater et al., 2019), Hypothesis 2 is formulated:
Vartanian, 2022). Such studies do not specifically state whether the
Hypothesis 2. Images of VIs have a positive impact on individuals’
images used are of SMIs. Therefore, given this research gap, the question
appearance anxiety.
of whether SMI images of elicit the appearance anxiety among social
media users needs further discussion. In the discussion of whether people react differently according to an
Moreover, with continuous development in science and technology, influencer’s nature (human vs. virtual), several researchers have
use of new SMI forms has been highlighted as a valuable emerging claimed that individuals’ cognition, emotion, behavior, and even related
marketing strategy (Farivar and Wang, 2022; Bu et al., 2022; Cheung brain activation regions present the same paradigm when encountering
et al., 2022; Appel et al., 2020). Virtual influencers (VIs), who are both types (i.e., VIs and HIs) (e.g., Stein et al., 2022). Social presence
designed to avoid the limitations of real humans, are believed to have a plays a significant role in the anthropomorphic interactions investigated
more prominent presence than human influencers on social media, in such studies (Potdevin et al., 2020). However, evidence from other
along with strong computing power and artificial intelligence (AI) al­ studies has identified the differences between the influence of computers
gorithms (Appel et al., 2020). Defined as “embodied virtual agents” (Tan and humans. People tend to tolerate emotional quirks when they face a
et al., 2020), several VIs have an anthropomorphized appearance and computer-generated partner and have greater involvement in conver­
are able to share attractive content, especially visual content, on social sations with human partners (Arsenyan and Mirowska, 2021). Further­
media (Arsenyan and Mirowska, 2021). Most VIs are built on their own more, people also had negative perceptions of VIs’ human-resembling
story lines and have distinct personalities. The interest generated by abilities (MacDorman, 2019; Stein et al., 2020). Most of these studies
these VIs among social media users is an aspect that needs further were dependent on the uncanny valley (UV) theory (Mousas et al.,
research. For example, millions of followers pay attention to the daily 2018), which describes the decreased affinity toward a virtual person
fashion choices and routines of Lil Miquela, the first computer-generated who increasingly resembles a human. Considering such contradictory
imagery (CGI) in the world. As VIs gain increasingly prominent social conclusions, the following research question (RQ) was formulated:
and marketing value, the influential mechanism of VIs has generated
RQ1. Is the effect of VI images different from that of HI images on
interest among many scholars (Arsenyan and Mirowska, 2021; Faddoul
individuals’ appearance anxiety?
and Chatterjee, 2020). Past studies have focused on people’s reactions to
VIs and human influencers (HIs) on social media and investiagted which
VIs’ characteristics that may impact people’s perceptions (e.g., Chiou 1.2. The mediating role of state appearance comparison
et al., 2019; Schroeder et al., 2020). Based on the above discussion, the
present study aims to examine whether images of HIs and VIs induce Comparison is significant in the mechanisms underlying the effects of
social media users’ appearance anxiety and whether there is a significant social media on individuals (Valle et al., 2021; Fardouly et al., 2017;
difference between the effects of HIs and VIs. Hawes et al., 2020). Decades ago, researchers noticed that the sharing of
Considering the growing number of studies of gender differences in an ideal life and a positive self-image on social media may induce
the relationships among social media and appearance-related outcomes negative feelings (such as jealousy) among others (Tandoc et al., 2015).
(e.g., Rogers and Rosseau, 2022), the present study focuses on the Further research demonstrated that social comparison can help in un­
impact of SMI images on young females as this group has been found to a derstanding this phenomenon (Saunders and Eaton, 2018), in particular
higher level of self-objectication, appearance-related motivations, and upward comparisons, which are characterized by comparing oneself to
social media appearance preoccupation (e.g., Zimmer-Gembeck et al., someone deemed as more attractive or superior in some way (McComb
2021; Rodgers et al., 2020; Choukas-Bradley et al., 2022). and Mills, 2022). According to the social comparison theory, individuals
evaluate their social value by comparing their performances and char­
1.1. The effect of SMI images on social media users’ appearance anxiety acteristics in various aspects (such as finance, appearance, and intelli­
gence) with those of others. Additionally, individuals are more likely to
Extensive experiments and surveys have been conducted to confirm perceive those who are similar to themselves as social comparison tar­
the relationship between exposure to ideal imagery and appearance gets. (Fardouly et al., 2017). As most of the content is generated by
anxiety (e.g., Slater et al., 2019). In this mechanism, an friends and peers, social media produces a higher level of perceived
appearance-related photo is associated with appearance comparison, proximity than traditional media (Rodgers et al., 2021). The platforms
which, in turn, elicits body dissatisfaction (Lee and Lee, 2021). It has provided by social media are also highly accessible and visual, facili­
been claimed that celebrities are particularly harmful targets of tating in social comparison in a relatively easy manner (Fardouly et al.,
appearance comparison due to their attractive appearance (Fardouly 2017). Consequently, people may tend to compare themselves with
et al., 2021). On social media, most HIs meet the general social figures on social media.
appearance ideals (i.e., thin, fair-complexioned, and young; Valle et al., Moreover, most images on social media are edited by modern media
2021). Their attractiveness and appearance characteristics have been technology (Cohen et al., 2018). User-friendly photo-editing tools meet
demonstrated to be the factors that influence social media users the most a user’s desire to perfect their preferred self-image. Thus, exposure to
(Fernandes et al., 2022; Sokolova and Kefi, 2019). Therefore, the first these curated images on social media, which are closer to the idealized
hypothesis of this study is: societal appearance, become upward comparison targets for social
media users (Anixiadis et al., 2019; Fardouly and Rapee, 2019).
Hypothesis 1. HI images have a positive impact on an individual’s
Therefore, stress, depression, and anxiety are triggered and intensified,
appearance anxiety.
when users realize that their appearance cannot meet the societal beauty
However, no research has identified the direct association between standards (Saunders and Eaton, 2018; Tiggemann and Zinoviev, 2019).
VIs’ appearance and social media users’ appearance anxiety. Re­ Studies have been conducted to identify the above-mentioned mecha­
searchers have demostrated the impact of virtual people’s appearance nism. Researchers have found that comparison plays an important
on individuals’ attitudes and behaviors (e.g., Hanus and Fox, 2015). mediating role in social media’s effect on individuals’
Visually attractive VIs are more likely to elicit social responses and appearance-related concerns (e.g., Carter and Vartanian, 2022; Boursier
behavioral change in humans (Arsenyan and Mirowska, 2021). As it is et al., 2020; Slater et al., 2019). The academic concept, state appearance
their conformity with attractive stereotypes which gives VIs their comparison, has been used in these studies to explain the extent to which
persuasive power and effectiveness (Faddoul and Chatterjee, 2020), the participants engaged in appearance comparison processing while
many VIs are designed to present an idealized appearance. Based on the viewing the images (e.g., McComb and Mills, 2022). State measure has

2
F. Deng and X. Jiang Journal of Retailing and Consumer Services 71 (2023) 103233

been identified as a useful tool to validate the power of experimental inducing individuals’ body dissatisfaction (Valle et al., 2021; Cohen
manipulations when specific image disturbances are used (Reed et al., et al., 2019). Second, Farivar and his colleagues (2021) claimed that
1991). Therefore, we speculate that individuals’ state appearance body satisfaction is greater when women are exposed to a favorite ce­
comparison may function as a mediator between SMI images and lebrity with whom they have a para-social relationship, and Brown and
appearance anxiety. Thus, the following hypothesis has been Tiggemann (2016) believed that high celebrity worship worsens body
formulated: dissatisfaction after exposure to thin ideal celebrity images. As a
newly-born VI, Ayayi has fewer than 60,000 followers (compared to the
Hypothesis 3. Appearance comparison mediates the effect between
3 million following Miquela, another VI). Referring to McComb and
the images of influencers (both HI and VI) and individuals’ appearance
Mills’ study (2022), Ayayi was selected to minimize the impact of ce­
anxiety.
lebrity worship and para-social relationships on appearance anxiety.
According to the above discussion of whether people react differently The second set of images was selected from an HI with almost the same
depending on an influencer’s nature (human vs. virtual), we formulate haircut, body shape, and number of followers as Ayayi (see Supple­
RQ2 as follows: mentary Materials for sample images). Ten images each were selected
from both Ayayi’s and the HI’s websites, including full body shots and
RQ2. Do images of VIs and HIs affect individuals’ state appearance
close-ups of their faces. No other people were shown in those images.
comparison differently?
The third set of images comprised scenic pictures including mountains,
Fig. 1 shows the research model of this study. the ocean, and gardens.
Physical attractiveness was examined on a seven-point Likert scale,
2. Method as in a previous study (Li and Peng, 2021). The scale consists of four
items and asks participants whether they feel that the woman in the
2.1. Experiment design picture is classy, beautiful, elegant, and sexy. As a pretest, 60 female
participants aged 18–35 were recruited and divided evenly into two
A single-factor inter-group experiment was conducted in this study. groups. Each group was shown one set of images. An independent
Three (HI group vs. VI group vs. control group) levels were created. The samples t-test showed no significant difference (MHI = 5.70, SD = 0.82;
participants were randomly assigned to the above groups evenly, and MVI = 5.50, SD = 0.99; t = 0.858, p > 0.05, df = 58) in attractiveness
they reviewed the corresponding images. between the two sets of images.

2.2. Participants 2.3.2. Measures

The participants (n = 202) were recruited through Credamo, an 2.3.2.1. State appearance anxiety. The measurement of a participant’s
Internet questionnaire platform in China, and were paid 5 RMB each for appearance anxiety was based on the state subscale of the physical
participation. Credamo contains a database of 2.8 million national appearance state-trait anxiety scale (PASTA-S) developed by Reed et al.
samples. Referring to the operation from Mulgrew and his colleagues’ (1991). The PASTA-S measures a person’s levels of anxiety, tension, or
study (2017), only 18–35-year-old (the age group most concerned about nervousness regarding their own appearance and specific body parts
their appearance) females were eligible to participate in this experi­ “right now” for 16 body parts (hips, tights, stomach, arms, waist, neck,
ment. The responses were excluded if the participants failed to answer chin, feet, legs, forehead, skin, curve, weight, height, wrists, and but­
two attention check questions (n = 5), spent too much or little time on tocks) on a five-point Likert scale (1 = not at all, 2 = slightly, 3 =
the experiment (n = 14), or chose the same answer for all the measures moderately, 4 = very much so, and 5 = exceptionally so). The PASTA has
(n = 5). Finally, 178 respondents were remained (60 for the VI group, 58 been widely used in research on body image (e.g., Bue, 2020; Porras-­
for the HI group, and 60 for the control group), with a mean age of 26.24 Garcia et al., 2020). The reliability of the scale was good (α = 0.909),
(SD = 4.82) and a mean body mass index (BMI) of 19.71 (SD = 3.18). and the total score of the 16 items represented the extent of each par­
ticipant’s appearance anxiety, with a higher score indicating a higher
2.3. Materials level of state appearance anxiety.

2.3.1. Stimuli materials 2.3.2.2. State appearance comparison. The state appearance comparison
Three sets of 10 images were used in the experiment. The first set scale developed by Tiggemann and McGill (2004) was adapted to
comprised images of Ayayi, the three-dimensional (3D) realistic virtual measure the level at which the participants compared their appearance
character created in 2021 by a Chinese Internet company. Ayayi shares with that of the individuals in the images. A four-item, seven-point
her wardrobe and makeup style realistically and energetically through Likert scale (from 1 = not at all to 7 = very much) was used. The first two
social media. At present, Ayayi is at the initial stage of becoming a items asked the participants to evaluate the level at which they thought
virtual fashion influencer, and she was selected for this study for two about (1) their whole appearance and (2) specific parts of their bodies
principal reasons.First, as a virtual fashion blogger Ayayi is designed to while observing the person in the images. The last two items asked the
have a young, slim, toned body, and an exquisite appearance. According participants to evaluate the level at which they compared (3) their
to previous studies, these images have the highest impact in terms of whole appearance and (4) specific body parts with those of the person in
the images. The total score of the four items represents the extent of each
participant’s appearance comparison, with a higher score indicating a
higher level of comparison with the person in the images (α = 0.925).

2.3.2.3. Covariates. Social media use (e.g., time and frequency) has
proven to have significant effects on individuals’ appearance anxiety
(Hawes et al., 2020; Phillips and Wisniewski, 2021). Moreover, BMI
(Carter and Vartanian, 2022) and para-social relationships (Farivar
et al., 2021) have been correlated with state body dissatisfaction. Thus,
these variables were measured as covariates in the following analysis
process. As regards, the social media use, the participants were asked the
Fig. 1. Research model.

3
F. Deng and X. Jiang Journal of Retailing and Consumer Services 71 (2023) 103233

average time they spent on social media every day (<1 h, 1–3 h, 3–5 h, SD = 9.85; t = − 4.408, p < 0.05, df = 116). Thus, Hypothesis 1 was
>5 h), and the frequency of their social media use per day (<10 times, supported. Participants who viewed VI images also reported a higher
10–20 times, 20–30 times, >30 times). The value of participants’ BMI level of state appearance anxiety than those who viewed control images
was calculated by dividing their weight (kg) by the square of their height (MVI = 35.15, SD = 11.31; MControl = 31.20, SD = 9.85; t = − 2.040, p <
(m). In addition, the participants were asked whether they were already 0.05, df = 118). Therefore, Hypothesis 2 was supported. In addition, the
aware of the influencer, results indicate that participants in the HI condition had higher
appearance anxiety, on average, than those in the VI condition (MHI =
2.4. Procedure 39.76, SD = 11.18; MVI = 35.15, SD = 11.31; t = 2.226, p < 0.05, df =
118). Thus, RQ1 was answered.
To not disclose the real purpose of this experiment, the researchers The mediating effect of the state appearance comparison was
asked the participants to judge whether the selected influencer was examined through bootstrapping analysis from the SPSS PROCESS. The
suitable as a social media spokesperson for a makeup brand for a specific data from the HI and VI group were analyzed with 5000 samples (Model
consumer group. First, the participants were asked to provide their 4; Hayes, 2017). BMI, frequency of social media use per day, and
personal information and social media usage situation. After being given average time spent on social media every day were also included in this
a brief written introduction about the definitions of a VI (VI group) and model as covariates, along with whether the participants were previ­
HI (HI group), the participants were asked to view each image for at ously aware of the influencer. As shown in Table 2, the total effect of
least 10 s and answer whether the person in them was suitable to influencers’ images on participants’ appearance anxiety was significant
endorse a makeup brand. Next, the participants had to finish the state (b = 4.532, SE = 2.116, 95% CI [0.340, 8.724]). In addition, the indirect
appearance anxiety scale, the state appearance comparison scale, and effect of image exposure on the participants’ appearance anxiety
the attractiveness scale. Participants in the control group were asked to through the state appearance comparison was significant (b = 2.787, SE
view 10 landscape images (for at least 10 s each) and then finish the state = 1.119, 95% CI [0.837, 5.250]). However, the direct effect of influ­
appearance anxiety scale and social media use situation items. The encers’ images on participants’ appearance anxiety was not significant
participants were rewarded 5 RMB after the researchers checked their (b = 1.744, SE = 1.940, 95% CI [− 2.101, 5.589]). Thus, Hypothesis 3
responses. was supported. An individual’s state comparison played a total medi­
ating role between exposure to the SMI images and their state appear­
2.5. Data analyses ance anxiety. The positive coefficients indicated that individuals who
engaged in greater appearance comparison experienced greater
Several checks were conducted prior to the main analysis process. appearance anxiety levels.
First, an independent samples t-test showed that there is no significant Based on the independent samples t-test, participants who viewed HI
difference in attractiveness (MHI = 6.07, SD = 0.61; MVI = 5.77, SD = images engaged in a higher level of state appearance comparison than
1.10; t = − 1.843, p > 0.05, df = 116) between the HI and the VI group. those who viewed VI images (MHI = 20.24, SD = 5.85; MVI = 17.67, SD
Second, a one-way ANOVA analysis showed that there is no difference = 6.53; t = 2.258, p < 0.05, df = 116). Therefore, RQ2 was answered.
among the three exposure groups in terms of age [F (2,175) = 1.568, p > HIs elicited a significantly higher state appearance comparison than VIs.
0.05], average time spent on social media every day [F (2,175) = 1.752,
p > 0.05], frequency of social media use per day [F (2,175) = 1.234, p > 4. Discussion
0.05], or BMI [F (2,175) = 0.532, p > 0.05]. For the following analysis,
the control condition was coded as 0, VI condition was coded as 1, and 4.1. Conclusions and implications
HI condition was coded as 2. Independent samples t-test analyses were
conducted to test Hypotheses 1 and 2, and to answer RQ1 and 2. In Based on the above analysis, we concluded from the present study
addition, the SPSS PROCESS was conducted to test Hypothesis 3. that 1) the randomization of the experiment procedure resulted in no
Bootstrapping analysis with 5000 samples (Model 4; Hayes, 2017) was differences among HI, VI, and scenic/control groups on the covariates;
used to examine the mediation effect of the state appearance 2) the groups of participants who viewed the images of VIs and HIs both
comparison. reported significantly higher scores in appearance anxiety than the
group which viewed scenic images; 3) participants who viewed VI’s
3. Results images reported significantly lower scores in state appearance anxiety
than those who viewed HI images; 4) the level of state appearance
A one-way ANOVA was conducted to test the significance of main comparison elicited by VIs was significantly lower than that elicited by
effect. State appearance anxiety [F (2,175) = 9.289, p < 0.001] and state HIs; and 5) individuals’ state appearance comparison plays a mediating
appearance comparison [F (2,175) = 5.078, p < 0.05] were significantly role in the association between an influencer’s images and the in­
different among exposure groups. According to the results of the inde­ dividual’s appearance anxiety. These findings provide evidence for the
pendent samples t-test shown in Table 1, participants exposed to HI correlation between influencers’ images and social media users’
images reported a higher level of state appearance anxiety than those appearance anxiety. Contrary to previous works claiming that in­
exposed to scenic images (MHI = 39.76, SD = 11.18; MControl = 31.20, dividuals’ emotional reaction when encountering VIs present the same
paradigm as that when encountering real HIs (e.g., Kramer et al., 2015),
this study argued that the level of social media users’ appearance anx­
Table 1
The Results of Independent Samples t-test. iety varies with the nature of the influencers (real human vs. virtual
human).
Dependent Images M SD t p df

State Appearance HI 39.76 11.18

Table 2
Anxiety Control 31.20 9.85 − 4.408 0.000 116
Total, direct, and indirect effects of images on appearance anxiety.
VI 35.15 11.31
Control 31.20 9.85 − 2.040 0.044 118 Effect Coefficient SE 95% Bootstrap CI
HI 39.76 11.18
Lower Upper
VI 35.15 11.31 2.226 0.028 116
State Appearance HI 20.24 5.85 Total 4.532 2.116 0.340 8.724
Comparison VI 17.67 6.53 2.258 0.026 116 Direct 1.744 1.940 − 2.101 5.589
Indirect 2.787 1.119 0.837 5.250
Notes: HI means human influencer, VI means virtual influencer.

4
F. Deng and X. Jiang Journal of Retailing and Consumer Services 71 (2023) 103233

The findings of this study have several implications for influencer the human resemblance of VIs has an impact on the latter’s influence
endorsement marketing practice. First, with the rapid development of (Pelau et al., 2021). Human resemblance was earlier defined as an in­
SMI endorsement marketing, our findings have implications for practi­ dividual’s perception of humanistic features in objects (Pelau et al.,
tioners using SMIs for marketing. The effect of human influencers on 2021), and it is believed to be a significant variable that can affect in­
consumers’ appearance anxiety should be considered when marketers dividuals’ reaction to AI-associated technologies, including VIs (Arsen­
specify strategies to stimulate consumers’ purchase intention online. For yan and Mirowska, 2021; Kim et al., 2022). The boundary between HIs
instance, the effect of an HI’s endorsement of cosmetic products may be and VIs is becoming increasingly blurred. As VIs increasingly resemble
better than that of a VI’s endorsement, because the appearance anxiety real humans, further research is required to determine whether they will
aroused by SMIs may cause a stronger impulse among consumers to buy induce the same appearance anxiety as HIs do. After all, according to the
compensation products. UV hypothesis (Arsenyan and Mirowska, 2021), a VI being particularly
Second, the application of VI might be a useful strategy to protect similar to a real person will lead to an extreme emotional aversion
individuals from developing negative appearance concerns due to social among humans.
media use (e.g., Song et al., 2014; Giorgianni et al., 2020). Even though The second direction involves the discussion of the effects of users’
the distance between virtual people and real humans is increasingly characteristics. To date, research has shown that the appearance com­
reducing due to the rapid development of AI technology, it may be parison tendency (Politte-Corn and Fardouly, 2020) and self-concept
advisable for practitioners to retain certain elements of “virtuality” clarity (Carter and Vartanian, 2022), of individuals play a moderating
while designing a virtual person. Researchers have identified that people role in the relation between social media exposure and the state
are more likely to compare themselves with individuals who are similar appearance comparison. However, the moderating effect of these vari­
to them (Trekels et al., 2018). Maintaining the difference between VIs ables were not significant in this study’s pretest experiments (not re­
and real humans can reduce the appearance comparison between social ported in the article due to their non-significant effect). Thus, research
media users and VIs, thus control ling the appearance anxiety level of on more possible moderating variables is needed. For example, it has
users. been proposed that media literacy skills have a protective effect by
However, when a VI is used as the spokesperson of brands and minimizing social media’s negative impact on individuals’ appearance
products, we must also take into account that they are not fully identi­ anxiety (Rodgers et al., 2021; Tamplin et al., 2018). Whether media
fiable as real people. The product category of VI endorsement needs literacy skills play a moderating role in the context of new media
careful consideration, because a virtual human is not as effective as a technology, such as virtual figures based on AI, needs further research.
real human in specific ways. For instance, we believe that virtual people Third, the emotional relationship between consumers and VIs, such
can “see” the world like real people but cannot “taste” food like real as worship and para-social relationships, deserve further discussion
people. (Farivar et al., 2021). Based on the results of our experiment, the par­
ticipants’ previous awareness of this influencer (b = 3.532, SE = 1.147,
4.2. Limitations 95% CI [1.258, 5.805]) was also identified to have a positive effect on
their state appearance comparison. Future research can further explore
This study inevitably has some limitations. Most previous studies how these emotional relationships affect consumers’ psychological
used images of several different people from social media (Valle et al., variables and purchase intentions.
2021). However, by limiting this study to the number of followers, only Public comments on social media also play an important role in
one influencer was selected for each group. This may have affected the marketing and social media’s effect on individuals’ body images
validity of the experiment, or may have increased the impact of the (Arsenyan and Mirowska, 2021; Hawes et al., 2020; Valle et al., 2021).
influencer herself. Second, the effects of variables such as a para-social Arsenyan and Mirowska (2021) predicted that the ideal images of SMIs
relationships and celebrity worship were not perfectly avoided by con­ might be more damaging to individuals’ body images in a higher-risk
trolling the number of followers. Some participants were aware of the one (comments praising the person’s appearance) than a lower-risk
influencers selected in this study. Therefore, if sufficient research funds context (comments highlighting the unrealistic nature of the images).
are available, the ideal research design would be to invite the technology Future studies could examine the role of other users’ comments in the
company involved to create a brand-new VI. Third, it is difficult when image-anxiety mechanism. Further, different media forms (words, pic­
using an online research platform such as Credamo to contact partici­ tures, and videos) may have different effects on individuals’ appearance
pants for further interview. Future research may consider the combi­ anxiety. Research involving in-depth discussion of the differences
nation of online and offline experiments. among text-based media (e.g., Twitter), picture-based media (e.g.,
Instagram), and short video-based media (e.g., TikTok) is needed in the
4.3. Future research directions future.

Considering the increasing number of studies on VIs’ impact on so­ Data availability
cial media users, an increasing amount of research should be conducted
to gain further understanding of this mechanism. One direction includes Data will be made available on request.
the characteristics of VIs, for instance, whether people’s perception of

Appendix B. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.jretconser.2022.103233.

Appendix A. Part of the Experiment Stimuli

Part of VI images:

5
F. Deng and X. Jiang Journal of Retailing and Consumer Services 71 (2023) 103233

Part of HI images:

References discrepancy. J. Bus. Res. 132, 146–157. https://doi.org/10.1016/j.

jbusres.2021.04.025.
Boursier, V., Gioia, F., Griffiths, M.D., 2020. Do selfie-expectancies and social appearance
Anixiadis, F., Wertheim, E.H., Rodgers, R., Caruana, B., 2019. Effects of thin-ideal
anxiety predict adolescents’ problematic social media use? Comput. Hum. Behav.
Instagram images: the roles of appearance comparisons, internalization of the thin
110 https://doi.org/10.1016/j.chb.2020.106395. Article 106395.
ideal and critical media processing. Body Image 31, 181–190. https://doi.org/
Brown, Z., Tiggemann, M., 2016. Attractive celebrity and peer images on Instagram:
10.1016/j.bodyim.2019.10.005.
effect on women’s mood and body image. Body Image 19, 37–43. https://doi.org/
Appel, G., Grewal, L., Hadi, R., Stephen, A.T., 2020. The future of social media in
10.1016/j.bodyim.2016.08.007.
marketing. J. Acad. Market. Sci. 48, 79–95. https://doi.org/10.1007/s11747-019-
Bu, Y., Parkinson, J., Thaichon, P., 2022. Influencer marketing: homophily, customer
00695-1.
value co-creation behaviour and purchase intention. J. Retailing Consum. Serv. 66
Arsenyan, J., Mirowska, A., 2021. Almost human? A comparative case study on the social
https://doi.org/10.1016/j.jretconser.2021.102904. Article 102904.
media presence of virtual influencers. Int. J. Hum. Comput. Stud. 155 https://doi.
Bue, A.C.C., 2020. The looking glass selfie: Instagram use frequency predicts visual
org/10.1016/j.ijhcs.2021.102694. Article 102694.
attention to high-anxiety body regions in young women. Comput. Hum. Behav. 108
Aw, E.C.-X., Chuah, S.H.-W., 2021. “Stop the unattainable ideal for an ordinary me!”
https://doi.org/10.1016/j.chb.2020.106329. Article 106329.
Fostering parasocial relationships with social media influencers: the role of self-

6
F. Deng and X. Jiang Journal of Retailing and Consumer Services 71 (2023) 103233

Carter, J.J., Vartanian, L.R., 2022. Self-concept clarity and appearance-based social Mousas, C., Anastasiou, D., Spantidi, O., 2018. The effects of appearance and motion of
comparison to idealized bodies. Body Image 40, 124–130. https://doi.org/10.1016/ virtual characters on emotional reactivity. Comput. Hum. Behav. 86, 99–108.
j.bodyim.2021.12.001. https://doi.org/10.1016/j.chb.2018.04.036.
Cheung, M.L., Leung, W.K.S., Aw, E.C., Koay, K.Y., 2022. I follow what you post!”: the Mulgrew, K.E., Stalley, N.L., Tiggemann, M., 2017. Positive appearance and functionality
role of social media influencers’ content characteristics in consumers’ online brand- reflections can improve body satisfaction but do not protect against idealised media
related activities (COBRAs). J. Retailing Consum. Serv. 66 https://doi.org/10.1016/ exposure. Body Image 23, 126–134. https://doi.org/10.1016/j.
j.jretconser.2022.102940. Article 102940. bodyim.2017.09.002.
Choukas-Bradley, S., Roberts, S.R., Maheux, A.J., Nesi, J., 2022. The perfect storm: a Naeem, M., 2021. Do social media platforms develop consumer panic buying during the
developmental-sociocultural framework for the role of social media in adolescent fear of Covid-19 pandemic. J. Retailing Consum. Serv. 58 https://doi.org/10.1016/j.
girls’ body image concerns and mental health. Clin. Child Fam. Psychol. Rev. 25, jretconser.2020.102226. Article102226.
681–701. https://doi.org/10.1007/s10567-022-00404-5. Pelau, C., Dabija, D.-C., Ene, I., 2021. What makes an AI device human-like? The role of
Chiou, E.K., Schroeder, N.L., Craig, S.D., 2019. How we trust, perceive, and learn from interaction quality, empathy and perceived psychological anthropomorphic
virtual humans: the influence of voice quality. Comput. Educ. 146 https://doi.org/ characteristics in the acceptance of artificial intelligence in the service industry.
10.1016/j.compedu.2019.103756. Article 103756. Comput. Hum. Behav. 122 https://doi.org/10.1016/j.chb.2021.106855. Article
Cihan, A., Cihan, E., 2019. Interrelation between appearance anxiety and sexual 106855.
functions in women: the role of surgical scars, morphologic features, and Phillips, W.J., Wisniewski, A.T., 2021. Self-compassion moderates the predictive effects
accompanying depression. J. Sex. Med. 16 (11), 1769–1778. https://doi.org/ of social media use profiles on depression and anxiety. Comput. Human Behav. Rep.
10.1016/j.jsxm.2019.08.004. 4 https://doi.org/10.1016/j.chbr.2021.100128. Article 100128.
Cohen, R., Fardouly, J., Newton-John, T., Slater, A., 2019. #BoPo on Instagram: an Politte-Corn, M., Fardouly, J., 2020. #Nomakeupselfie: the impact of natural no-makeup
experimental investigation of the effects of viewing body positive content on young images and positive appearance comments on young women’s body image. Body
women’s mood and body image. New Media Soc. 21 (7), 1546–1564. https://doi. Image 34, 233–241. https://doi.org/10.1016/j.bodyim.2020.07.001.
org/10.1177/1461444819826530. Porras-Garcia, B., Serrano-Troncoso, E., Carulla-Roig, M., Soto-Usera, P., Ferrer-
Cohen, R., Newton-John, T., Slater, A., 2018. ‘Selfie’-objectification: the role of selfies in Garcia, M., Figueras-Puigderrajols, N., Gutiérrez-Maldonado, J., 2020. Virtual reality
self-objectification and disordered eating in young women. Comput. Hum. Behav. body exposure therapy for anorexia nervosa. A case report with follow-up results.
79, 68–74. https://doi.org/10.1016/j.chb.2017.10.027. Front. Psychol. 11 https://doi.org/10.3389/fpsyg.2020.00956. Article 956.
Farivar, S., Wang, F., 2022. Effective influencer marketing: a social identity perspective. Potdevin, D., Sabouret, N., Clavel, C., 2020. Intimacy perception: does the artificial or
J. Retailing Consum. Serv. 67 https://doi.org/10.1016/j.jretconser.2022.103026. human nature of the interlocutor matter. Int. J. Hum. Comput. Stud. 142 https://doi.
Article 103026. org/10.1016/j.ijhcs.2020.102464. Article 102464.
Farivar, S., Wang, F., Yuan, Y., 2021. Opinion leadership vs. para-social relationship: key Reed, D.L., Thompson, J.K., Brannick, M.T., Sacco, W.P., 1991. Development and
factors in influencer marketing. J. Retailing Consum. Serv. 59 https://doi.org/ validation of the physical appearance state and trait anxiety scale (PASTAS).
10.1016/j.jretconser.2020.102371. Ariticle 102371. J. Anxiety Disord. 5 (4), 323–332. https://doi.org/10.1016/0887-6185(91)90032-o.
Fardouly, J., Pinkus, R.T., Vartanian, L.R., 2021. Targets of comparison and body image Rogers, R.F., Rosseau, A., 2022. Social media and body image: modulating effects of
in women’s everyday lives: the role of perceived attainability. Body Image 38, social identities and user characteristics. Body Image 41, 284–291. https://doi.org/
219–229. https://doi.org/10.1016/j.bodyim.2021.04.009. 10.1016/j.bodyim.2022.02.009.
Faddoul, G., Chatterjee, S., 2020. A quantitative measurement model for persuasive Rodgers, R.F., Paxton, S.J., Wertheim, E.H., 2021. #Take idealized bodies out of the
technologies using storytelling via a virtual narrator. Int. J. Hum. Comput. Interact. picture: a scoping review of social media content aiming to protect and promote
36, 1585–1604. https://doi.org/10.1080/10447318.2020.1768670. positive body image. Body Image 38, 10–36. https://doi.org/10.1016/j.
Fardouly, J., Rapee, R.M., 2019. The impact of no-makeup selfies on young women’s bodyim.2021.03.009.
body image. Body Image 28, 128–134. https://doi.org/10.1016/j. Rodgers, R.F., Mclean, S.A., Gordon, C.S., Slater, A., Marques, M.D., Jarman, H.K.,
bodyim.2019.01.006. Paxton, S.J., 2020. Development and validation of the motivations for social media
Fardouly, J., Pinkus, R.T., Vartanian, L.R., 2017. The impact of appearance comparisons use scale (MSMU) among adolescents. Adolescent Res. Rev. 6, 425–435. https://doi.
made through social media, traditional media, and in person in women’s everyday org/10.1007/s40894-020-00139-w.
lives. Body Image 20, 31–39. https://doi.org/10.1016/j.bodyim.2016.11.002. Saunders, J.F., Eaton, A.A., 2018. Snaps, selfies, and shares: how three popular social
Fardouly, J., Vartanian, L.R., 2016. Social media and body image concerns: current media platforms contribute to the sociocultural model of disordered eating among
research and future directions. Curr. Opin. Psychol. 9, 1–5. https://doi.org/10.1016/ young women. Cyberpsychol., Behav. Soc. Netw. 21 (6), 343–354. https://doi.org/
j.copsyc.2015.09.005. 10.1089/cyber.2017.0713.
Fernandes, T., Nettleship, H., Pinto, L.H., 2022. Judging a book by its cover? The role of Schroeder, N.L., Chiou, E.K., Craig, S.D., 2020. Trust influences perceptions of virtual
unconventional appearance on social media influencers effectiveness. J. Retailing humans, but not necessarily learning. Comput. Educ. 160 https://doi.org/10.1016/j.
Consum. Serv. 66 https://doi.org/10.1016/j.jretconser.2022.102917. Article compedu.2020.104039. Article 104039.
102917. Seek, V., Barker, G., 2022. Does #beauty have a dark side? Testing mediating pathways
Ge, J., Gretzel, U., 2018. Emoji rhetoric: a social media influencer perspective. J. Market. between engagement with beauty content on social media and cosmetic surgery
Manag. 34, 1272–1295. https://doi.org/10.1080/0267257X.2018.1483960. consideration. Body Image 42, 268–275. https://doi.org/10.1016/j.
Giorgianni, F.E., Danthinne, E.S., Rodgers, R., 2020. Consumer warning versus systemic bodyim.2022.06.013.
change: the effects of including disclaimer labels on images that have or have not Slater, A., Cole, N., Fardouly, J., 2019. The effect of exposure to parodies of thin-ideal
been digitally modified on body image. Body Image 34, 249–258. https://doi.org/ images on young women’s body image and mood. Body Image 29, 82–89. https://
10.1016/j.bodyim.2020.07.007. doi.org/10.1016/j.bodyim.2019.03.001.
Hawes, T., Zimmer-Gembeck, M.J., Campbell, S.M., 2020. Unique associations of social Sokolova, K., Kefi, H., 2019. Instagram and YouTube bloggers promote it, why should I
media use and online appearance preoccupation with depression, anxiety, and buy? How credibility and parasocial interaction influence purchase intentions.
appearance rejection sensitivity. Body Image 33, 66–76. https://doi.org/10.1016/j. J. Retailing Consum. Serv. 53 https://doi.org/10.1016/j.jretconser.2019.01.011.
bodyim.2020.02.010. Article 101742.
Hayes, A.F., 2017. Introduction to Mediation, Moderation, and Conditional Process Song, H., Kim, J., Lee, K.M., 2014. Virtual vs. real body in exergames: reducing social
Analysis: A Regression-Based Approach, second ed. Guilford Publications. physique anxiety in exercise experiences. Comput. Hum. Behav. 36, 282–285.
Hanus, M.D., Fox, J., 2015. Persuasive avatars: the effects of customizing a virtual https://doi.org/10.1016/j.chb.2014.03.059.
salesperson‫׳‬s appearance on brand liking and purchase intentions. Int. J. Hum. Stein, J.P., Breves, P.L., Anders, N., 2022. Parasocial Interactions with Real and Virtual
Comput. Stud. 84, 33–40. https://doi.org/10.1016/j.ijhcs.2015.07.004. Influencers: the Role of Perceived Similarity and Human-Likeness. New media &
Kim, J., Kang, S., Bae, J., 2022. Human likeness and attachment effect on the perceived Society. https://doi.org/10.1177/14614448221102900 published online.
interactivity of AI speakers. J. Bus. Res. 144, 797–804. https://doi.org/10.1016/j. Stein, J.P., Appel, M., Jost, A., Ohler, P., 2020. Matter over mind? How the acceptance of
jbusres.2022.02.047. digital entities depends on their appearance, mental prowess, and the interaction
Kramer, N.C., Rosenthal-von der Pütten, A.M., Hoffmann, L., Sundar, S.S., 2015. Social between both. Int. J. Hum. Comput. Stud. 142 https://doi.org/10.1016/j.
effects of virtual and robot companions. In: The Handbook of the Psychology of ijhcs.2020.102463. Article 102463.
Communication Technology, 1st ed. Wiley, pp. 137–159. doi:10.1002/ Tan, S.-M., Liew, T.W., Gan, C.L., 2020. Motivational virtual agent in e-learning: the roles
9781118426456.ch6. of regulatory focus and message framing. Info. Learn. Sci. 121 (1/2), 37–51. https://
Lee, M., Lee, H.-H., 2021. Social media photo activity, internalization, appearance doi.org/10.1108/ILS-09-2019-0088.
comparison, and body satisfaction: the moderating role of photo-editing behavior. Tamplin, N.C., McLean, S.A., Paxton, S.J., 2018. Social media literacy protects against
Comput. Hum. Behav. 114 https://doi.org/10.1016/j.chb.2020.106579. Article the negative impact of exposure to appearance ideal social media images in young
106579. adult women but not men. Body Image 26, 29–37. https://doi.org/10.1016/j.
Li, Y., Peng, Y., 2021. What drives gift-giving intention in live streaming? The bodyim.2018.05.003.
perspectives of emotional attachment and flow experience. Int. J. Hum. Comput. Tandoc Jr., E.C., Ferrucci, P., Duffy, M., 2015. Facebook use, envy, and depression
Interact. 37 (14), 1317–1329. https://doi.org/10.1080/10447318.2021.1885224. among college students: is Facebooking depressing? Comput. Hum. Behav. 43,
MacDorman, K.F., 2019. In the uncanny valley, transportation predicts narrative 139–146. https://doi.org/10.1016/j.chb.2014.10.053.
enjoyment more than empathy, but only for the tragic hero. Comput. Hum. Behav. Tiggemann, M., McGill, B., 2004. The role of social comparison in the effect of magazine
94, 140–153. https://doi.org/10.1016/j.chb.2019.01.011. advertisements on women’s mood and body dissatisfaction. J. Soc. Clin. Psychol. 23
McComb, S.E., Mills, J.S., 2022. The effect of physical appearance perfectionism and (1), 23–44. https://doi.org/10.1521/jscp.23.1.23.2699179.
social comparison to thin-, slim-thick-, and fit-ideal Instagram imagery on young Tiggemann, M., Zinoviev, K., 2019. The effect of #enhancement-free Instagram images
women’s body image. Body Image 40, 165–175. https://doi.org/10.1016/j. and hashtags on women’s body image. Body Image 31, 131–138. https://doi.org/
bodyim.2021.12.003. 10.1016/j.bodyim.2019.09.004.

7
F. Deng and X. Jiang Journal of Retailing and Consumer Services 71 (2023) 103233

Trekels, J., Ward, L.M., Eggermont, S., 2018. I “like” the way you look: how appearance- Wahab, H.K.A., Tao, M., Tandon, A., Ashfaq, M., Dhir, A., 2022. Social media celebrities
focused and overall Facebook use contribute to adolescents’ self-sexualization. and new world order. What drives purchasing behavior among social media
Comput. Hum. Behav. 81, 198–208. https://doi.org/10.1016/j.chb.2017.12.020. followers? J. Retailing Consum. Serv. 68, 103076 https://doi.org/10.1016/j.
Valle, M.K., Gallego-Garcia, M., Williamson, P., Wadw, T.D., 2021. Social media, body jretconser.2022.103076.
image, and the question of causation: meta-analyses of experimental and Zimmer-Gembeck, M.J., Hawes, T., Pariz, J., 2021. A closer look at appearance and
longitudinal evidence. Body Image 39, 276–292. https://doi.org/10.1016/j. social media: measuring activity, self-presentation, and social comparison and their
bodyim.2021.10.001. associations with emotional adjustment. Psychol. Popular Media 10 (1), 74–86.
Velkoff, E.A., Smith, A.R., 2020. Preliminary development of an implicit association test https://doi.org/10.1037/ppm0000277.
to measure body dissatisfaction and predict disordered eating behaviors. Body Image
34, 51–58. https://doi.org/10.1016/j.bodyim.2020.05.005.