ArticlePDF Available

AI-Driven Influencer Marketing: Comparing the Effects of Virtual and Human Influencers on Consumer Perceptions

Authors:
  • Macromedia University of Applied Sciences

Abstract

Computer-generated virtual influencers are currently one of the most important brand communication trends driven by artificial intelligence (AI). While numerous studies on human social media influencers already exist, the field of virtual influencers is still largely unexplored, which is especially true regarding their impact on consumer perceptions. Against this background, the aim of this paper is to empirically investigate consumer perceptions of virtual influencers in comparison with traditional social media influencers. We conduct an exploratory experiment to test the effect of virtual and human influencers on perceived credibility, competence and likeability as well as on purchase intentions. The results show no significant differences between virtual and human influencers, except for the variable likeability. Implications for management and future research are discussed.
Copyright: Henry Stewart Publications 1
This is the authors’ manuscript of the following publication:
Böhndel, M., Jastorff, M., & Rudeloff, C. (2023). AI-driven influencer marketing:
Comparing the effects of virtual and human influencers on consumer perceptions. Journal
of AI, Robotics & Workplace Automation, 2(2), 165-174.
AI-driven influencer marketing: Comparing the effects of virtual and
human influencers on consumer perceptions
Abstract
Computer generated virtual influencers are currently one of the most important brand
communication trends driven by artificial intelligence. While numerous studies on human
social media influencers already exist, the field of virtual influencers is still largely
unexplored, which is especially true regarding their impact on consumer perceptions.
Against this background, the aim of this study is to empirically investigate consumer
perceptions of virtual influencers in comparison to traditional social media influencers.
We conduct an exploratory experiment to test the effect of virtual and human influencers
on credibility, competence, likability, and purchase intentions. The results show no
significant differences between virtual and human influencers, except for the variable
likeability. Implications for management and future research are discussed.
Keywords
virtual influencers, cgi influencers, AI-driven influencers, social media influencers, SMI,
influencer marketing, consumer perceptions, purchase intentions
Copyright: Henry Stewart Publications 2
AI-driven influencer marketing: Comparing the effects of virtual and
human influencers on consumer perceptions
Introduction
The implementation of social media as “web-based services that allow individuals to (1)
construct a public or semi-public profile within a bounded system, (2) articulate a list of
other users with whom they share a connection, and (3) view and traverse their list of
connections”1 in corporate and brand communication has become standard for companies
in recent years2,3,4. However, social media networks such as Instagram are often
overloaded with content, making it increasingly difficult for brands to attract consumers'
attention. In this context, the phenomenon of social media influencers (SMI) has
developed. SMI can be understood as social media users who, due to their reach on social
media networks, act as third-party endorsers for brands5.
In the context of influencer marketing, brands enter collaborations with SMI to
improve the success of their brand communication. Once a niche movement, influencer
marketing is estimated to be a $16.4 billion industry in 2022, with more than 75% of
advertisers intending to dedicate a budget to influencer marketing in 20236. In 2019
approximately 50% of internet users followed at least one influencer account on social
media and 40% indicated that they had bought a product after seeing it on Instagram or
YouTube7.
However, influencer marketing is also subject to constant change. Since 2020, a
new trend has been growing in the industry, which is different from the previous state of
influencer marketing: The emergence of virtual influencers (VI), who operate on the basis
of computer-generated imagery and artificial intelligence. Sands et al. define VI as "an
entity - humanlike or not - that is autonomously controlled by artificial intelligence and
visually presented as an interactive, real-time rendered being in a digital environment"8.
Comparable to SMI, VI have already been used by brands in the context of (virtual)
influencer marketing. For example one of the largest VI Miquela Sousa (@lilmiquela)
has already worked with several different fashion brands or tech brands like Samsung
where she resembled the embodiment of the campaign’s slogan “Do What You Can’t”9.
While a considerable body of research already exists on SMI7, VI can be
considered as unexplored compared to their human equivalent10.
Within the literature in which VI are mentioned, they are often discussed on a
theoretical or conceptual level. In addition, first empirical studies exist using e. g. the case
Copyright: Henry Stewart Publications 3
study approach, but so far there are only very few experimental studies that examine the
effect of VI on variables that are relevant in the context of influencer marketing and brand
communication. Consequently, the need for a more in-depth consideration of the
emerging VI trend is high. In particular, comparing SMI and VI seems interesting, as
companies always prefer the most efficient communication option and VI promises
efficiency advantages in influencer marketing, e.g. through lower costs and easier
handling of the influencers.
From this starting point, the objectives of this study can be derived. Since previous
research has focused on SMI, this work aims to add value to VI research. The focus of
the work is on the way influencers are perceived by consumers as relevant target groups
of brands. The differences in how consumers view VI and SMI will be explored.
Literature review
Social media influencers as opinion leaders
Based on the findings of Lazarsfeld et al. in the “Peoples Choice” Katz & Lazarsfeld
developed the concept of the "Two-Step-Flow of Communication" early on, which is
presented in detail in their work "Personal Influencer"11,12. According to the model, the
dissemination of information through mass media takes place in two steps. First, opinion
leaders receive the information from the mass media. In the second step, this information
as well as the personal interpretation and opinion of the opinion leaders is disseminated
to the population and the masses.
Building on this foundation, various other "flow of communication" models have
been developed. These include the "One-Step-Flow of Communication" and also various
"Multistep-Flow of Communication" models13 .
The principle of opinion leadership identified by Lazarsfeld & Katz12 can also be
applied nowadays. Prominent influencers on Instagram can address many people due to
their high reach and take on an important role in a networked and digitalised world. In
this context, they often also act as role models. Accordingly, they can also influence
opinions on brands and, for example, increase the intentions to buy products, provided
the influencers are perceived as credible, competent and likeable by the recipients7.
In this context, influencer marketing represents the commercial use of the opinion
leader concept by companies and brands. Leung et al. define influencer marketing as "a
strategy in which a firm selects and incentivizes online influencers to engage their
followers on social media in an attempt to leverage these influencers' unique resources to
promote the firm's offerings, with the ultimate goal of enhancing firm performance"14.
Copyright: Henry Stewart Publications 4
The main goals of influencer marketing activities in a commercial context are to increase
the value of brand messages and to ultimately influence the purchase behaviour of
consumers7.
Differentiation of influencers
Within the influencer market, various types of influencers can be identified. The current
state of research provides different approaches for this. E. g. one way of differentiating
influencers is the size of their reach. Influencers can be assigned to three different reach
sizes: Nano, Micro and Mega-influencers15 .
Nano-influencers are the smallest category and have only a few hundred
followers. Compared to the other types, they often have a high level of personal
identification and interaction with their followers and, accordingly, a high level of
credibility. Micro-influencers have a following in the four- to five-figure range. This type
of influencers is usually an expert in a niche or has a local connection. Consequently, they
are of high interest for small or medium-sized companies, for example, and still have a
high level of credibility. Mega-influencers have followers of several hundred thousand or
even millions of people. Accordingly, they reach a broad and often diverse target group.
Compared to the smaller influencer types, however, they have a lower interaction rate and
credibility due to their reach.
In addition to reach size, influencers can also be divided into different thematic
categories, depending on which product category they promote particularly intensively.
Furthermore, there are numerous other typologies that take into account, among other
things, the communication behaviour of influencers with their followers16. At the same
time, it becomes clear that the state of research to date relates almost exclusively to SMI.
The current development of VI has so far received little to no attention in the literature,
also with regard to the differentiation of influencers. In the following, therefore, a
differentiation of SMI and VI is given on the basis of the literature to date.
Virtual influencers
Sands et al. emphasise that VI are autonomously controlled by artificial intelligence and
visually presented as an interactive, real-time rendered being in a digital environment8.
From other authors, they are also coined as computer generated imagery-influencers. The
Copyright: Henry Stewart Publications 5
term computer generated imagery originates from the film industry and signifies 3D
computer animations10.
Accordingly, a VI is not a real person like a SMI. A VI is a purely virtual entity
that has been designed and created by agencies. Here, designers, social media managers
and programmers work together and give the VI attributes and character traits that are
more or less similar to a real person. Comparable to SMI, VI appear on social media
platforms such as Instagram and collaborate with brands there, report on their “lives” or
communicate opinions and recommendations on current topics10.
Contributions published by VI are controlled to varying degrees by algorithms.
The entire background story of the VI is fictitious and can be deliberately crafted to appeal
to a specific target group17. Thus, VI can also appear activist and, for example, advocate
for the "Black Lives Matter'' movement or embody different lifestyles such as veganism.
An example of this behaviour is the VI @noonoouri which dedicated a whole Instagram
Story Series to the Pride Month18.
In terms of influencer marketing, VI offer several advantages over SMI. For
example, they speed up the entire creative process of content creation17. In addition, CGI
influencers cannot get sick, they do not age and they do not have to be elaborately made
up before a shoot. Nor are their appearances or shoots linked to external circumstances
such as the weather. They are available at any time and can be a good presentation for
advertising brands17.
VI can be divided into two different categories. Category 1 is made up of VI that
are modelled on human likeness. They try to imitate the human appearance very closely
and are sometimes hardly distinguishable from a human influencers at first glance.
Miquaela Sousa is one of the most popular VI of this kind and has already worked with
many different brands such as Samsung or Calvin Klein.
The second category of VI are the "unique" VI. These do not aim to imitate the human
appearance, but represent a unique, virtual character. On the one hand, this can be human-
like, but on the other hand, it can also be a non-human avatar. One of the best-known
examples is @noonoouri on Instagram, which was conceived and created by the German
graphic designer Joerg Zuber in cooperation with his creative agency Opium Effect in
Copyright: Henry Stewart Publications 6
Munich. In the course of her career as an influencer, she has also worked with brands
such as Dior and Versace and is considered one of Germany's most famous VI.
Success factors of SMI
Based on the literature review, several success factors can be identified that positively
influence the impact of SMI in marketing and brand communication. According to the
current state of research, the variables perceived credibility, expertise and likeability are
decisive for the success of influencers7. These variables are briefly explained in the
following.
The concept of credibility has played a crucial role in persuasion research and
particularly in research on the effects of SMI19 . Central to the attribution of credibility is
the extent to which the communicator provides the correct and relevant information on a
subject matter from the perspective of the recipients. Furthermore, trustworthiness is part
of the construct of credibility19.
Another important variable is expertise. Expertise describes the knowledge and
intellectual abilities of individuals whose performance in a particular field is considered
to be above average20. In the context of this study, it can be assumed that SMI can exert
influence on their recipients, which is reinforced by their status as an expert. Therefore,
one aspect that determines the influence of SMI is their expert status.
Furthermore, likeability describes a positive emotional attitude towards another
person, which prevails due to certain similarities or affinities21. Consequently, likeability
towards influencers is largely determined by the recipient's identification with the
influencers. In case of asymmetrical proximity, this is also referred to as parasocial
interaction10. For credibility, likeability also plays a decisive role22.
Against this background, the following research questions can be formulated:
RQ1: How does consumer’s perception of credibility differ between VI and SMI?
RQ2: How does consumer’s perception as an expert differ between VI and SMI?
RQ3: How does consumer’s perception of likeability differ between VI and SMI?
As the state of research shows, consumers' purchase intentions are also positively
influenced by SMI23. Since increasing purchasing intentions is at the same time an
important goal of brands' influencer marketing activities, this study also compares the
Copyright: Henry Stewart Publications 7
effect of VI and SMI regarding this variable. Therefore, the following research question
can be formulated:
RQ4: How does the influence on consumer’s purchase intentions differ between VI and
SMI?
Empirical study
Study design
To investigate the effects of VI and SMI a laboratory experiment was conducted. In
preparation for the laboratory experiment, the different experimental groups were first
defined. Since the influence of the influencer type on the recipient's perceptions is to be
examined, the influencer type was defined as an independent variable. Consequently, the
effect of the influencer type on the following dependent variables is examined:
Credibility, expertise, likeability and purchase intentions.
The experimental groups were designed so that each experimental group was
shown only one influencer type. Consequently, experimental group 1 evaluated their
perception of VI and experimental group 2 their perception of SMI. By limiting the
number of influencer types to one per experimental group, confounding factors were
reduced. For example, a possible confounding factor when showing influencer types in
both experimental groups would be the order in which the two types are shown.
To measure the evaluation of dependent variables by the test persons, an online
questionnaire, incl. 5-point Likert scales, was created. Items were developed based on the
literature review.
Study participants were recruited via convenience sampling. To ensure the validity of the
results, participants were randomly distributed to the different groups. A total of 63
subjects participated in the experiment.
Treatment
Image and video posts by influencers of the two influencer types were chosen as stimuli.
These were presented to the test group. The structure of the presentation was completely
standardised. After an introduction to the research topic, the test groups were shown one
male and one female influencer of the specific influencer type. A differentiation of the
gender of the influencer was made in order to determine any effects of the gender of the
Copyright: Henry Stewart Publications 8
influencer on the perception of the recipients and to prevent a possible bias of the results
due to an unequal gender of the influencer.
Each chosen influencer was introduced with a short intro text about him or herself
and presented with three images and three videos, which were taken from the respective
Instagram profile. It should be noted that it was not communicated during the presentation
of the VI that they were virtual persons. This procedure is intended to prevent biasing
participants’ reaction to the stimulus.
Influencers were selected for the stimulus presentation according to defined
criteria based on the literature review. Care was taken to ensure that the influencers were
comparable. For this purpose, the types of influencers presented in the course of the
literature review were used as selection criteria. Accordingly, influencers with
comparable reach were sought in order to obtain reliable results. The classification into
nano-, micro- and mega-influencers was used for this purpose.
Since VI are a comparatively new development in influencer marketing, there are
only a few areas in which they are active. VI are primarily active in the fashion/lifestyle
sector. For this reason, only influencers from this area were selected for the stimulus
presentation. In addition, the portfolio of integrated influencers consists exclusively of
mega-influencers. Table I shows a list of the influencers integrated in the experiment.
[ Insert Table I around here ]
Pretest
In order to test the quality of the measurement and the feasibility of the experiment, a pre-
test was conducted before the data collection. For the pre-test, the procedure of the
experiment was run through with five participants, recruited via convenience sampling.
Minor adjustments in the procedure of the experiment, which were remarked by the
participants of the pre-test, resulted. The suitability of the questionnaire was confirmed.
Results
Table II gives an overview of the descriptive results of the study. In addition, independent
t-tests were conducted to measure whether significant differences exist between the
experimental groups regarding their perception of the different types of influencers.
[ Insert Table II around here ]
Copyright: Henry Stewart Publications 9
Credibility
The variable credibility consists of three items. Due to the alpha value, the respondents'
statements for the three items were examined together (see Table II).
With regard to the distribution of responses, it can be observed that subjects of the
VI group (32%) and the SMI group (33%) similarly often indicated a value of < 3 for the
three items. However, differences were observed in the extent of the positive evaluation.
While 14% of the test persons of the VI group gave a rating of 5, only 4% of the test
persons of the SMI group rated this statement as 5. Both groups similarly often gave a
value of 4 (VI=34%, SMI=30%) and a value of 3 (VI=29%, SMI=22%).
For further analyses, the data were summarised into a mean score. Subsequently,
the t- test revealed a p-value of .182. With regard to the significance level (0.05), it can
therefore be concluded that the determined p-value is above the defined significance level
(0.182>0.05). Consequently, no significant differences can be found between the mean
values of the VI group (M=3.026) and the SMI group (M=3.188). The t-value is -0.91
with a degree of freedom of df=113.
Expertise
The variable expertise also consists of three items that can be considered together due to
the alpha value (see Table II). For the examination of the dependent variable, the data
was also combined into a mean score.
When looking at the distribution of answers of the test persons, it can be observed
that a large part of the VI group (37%) rated the influencer with a value of 3. In
comparison, only 19% of the respondents in the SMI group gave a value of 3.
Respondents in the VI group tended to give a more negative rating. 30% of the SMI group
gave a value < 3. In the VI group, only 20% gave a value < 3. With regard to the more
positive evaluation, 34% of the VI group and 28% of the SMI group rated the influencer
with a value of 4. The high proportion of test persons in SMI group who gave a value of
5 (19%) is also striking.
The variable expertise was also tested for significant differences using the one-
sided t-test for independent samples. A p-value of .369 resulted. The significance level
here is below the determined threshold (0.369>0.05). Consequently, no significant
Copyright: Henry Stewart Publications 10
differences can be found between the mean values of the VI (M=3.34) and the SMI group
(M=3.4). The t-value is -0.34 with a degree of freedom of df=111.
Likeability
The dependent variable likeability also consists of three items. The calculation of the
dependent variable is also based on the mean scores determined (see Table II).
In terms of response distribution, 55% of the VI group gave the influencers a score of <
3. In the SMI group, 46% of responses accounted for a score of < 3. In addition, 24% of
the SMI group subjects gave a score of 4. In comparison, only 15% of the VI group gave
a score of 4.
A one-sided t-test for independent samples was also carried out for the variable
likeability. A p-value of .0499 resulted. Accordingly, a significant difference between the
mean values of VI group (M=2.46) and the SMI group (M=2.76) was found. The t-value
is -1.66 with a degree of freedom of df=118. Due to the significant result, Cohen's d was
calculated. This resulted in a value of d=0.3.
Purchase intentions
The last dependent variable considered consumer’s purchase intentions. In response to
the statement "I can imagine buying products advertised by the influencer if they are
relevant to me”. The SMI group respondents gave a value > 3 more often (31%) than the
VI group respondents (20%). A large proportion of the VI group respondents (31%) rated
the statement as 3. The SMI group respondents rated the statement as 1 more often than
the VI group respondents (27%>21%).
The comparison of the test groups displays a p-value of .41. The determined p-
value is outside the significance level (0.41>0.05). For this reason, it can be assumed that
there are no significant differences between the mean values of the VI group (M=2.55)
and the SMI group (M=2.6). The t-value is -0.23 with a degree of freedom of df=124.
Discussion
This study expands the research in the field of influencer marketing. There has been little
research on VI to date, therefore this study contributes to a better understanding of this
emerging trend. Overall, it can be observed that the mean values of the VI group tend to
be slightly lower than the mean values of the SMI group. Accordingly, VI tended to be
assessed more negatively overall than SMI. However, the differences are not significant
Copyright: Henry Stewart Publications 11
except for the variable likeability. Therefore, the results confirm Ahn et al. and Stein et
al. who also didn’t find significant differences in the perception of VI and SMI10,24.
Regarding RQ1, no significant differences could be found for the variable credibility. A
closer look at the mean values also shows that the differences between the two
experimental groups is minimal. For the VI group a mean of 3.03 was found (SMI group:
mean = 3.19). If one compares the distributions of the answers, only minor differences in
the positive evaluations can also be found. Due to the small difference in the mean values
and the marginal differences in the distribution of responses, it can be assumed that there
are no differences between the two experimental groups with regard to the dependent
variable credibility.
Regarding RQ2, no significant differences could be found either. A closer look at
the mean values of the two experimental groups also shows that they are very close to
each other. The VI group has a mean value of 3.34, while the SMI group has a mean value
of 3.4. Accordingly, no difference can be observed between the two experimental groups
regarding the dependent variable expertise. Consequently, it can be assumed that VI and
SMI are frequently and similarly perceived as experts.
When examining the dependent variable likeability (RQ3), on the other hand, a
significant difference was found between the groups. The calculation of the effect strength
using Cohen's d resulted in a value of d=0.3, which indicates that the effect strength is
rather weak. Nevertheless, it can be stated that there are clear differences between the test
groups with regard to the perception of likeability. Since the mean value of the VI group
(2.46) is lower than that of the SMI group (2.76), it can be deduced that VI are perceived
as likeable less often and less strongly than SMI. In other words, in terms of likeability,
VI are perceived significantly more negatively than SMI.
Finally, with regard to the dependent variable purchase intentions (RQ4), again
no significant differences could be found between the groups. Here too, the mean values
of the VI group (2.55) and SMI group (2.6) are very close to each other. When looking at
the distribution of the answers, no relevant differences could be found either. Overall,
however, it can be observed that purchase intentions were rated generally low in both
groups.
Managerial implications
All in all, given that respondents didn’t perceive significant differences between VI and
SMI (except for the variable likeability), the potential of VI for brand communication can
be approved. Therefore, it can be proposed that the technological innovation of Ai-driven
Copyright: Henry Stewart Publications 12
VI represents a trend that will probably not lose its relevance and will be increasingly
implemented by influencer marketing campaigns in the future.
It can also be expected that concepts such as virtual reality, augmented reality will
make VI even more relevant. This is especially true against the backdrop of the metaverse,
which supposedly represents a completely new way of digital life and digital
communication. VI may gain in importance, as SMI would also have to adopt an avatar
in the digital metaverse. Therefore, further research and investigations will be necessary
to be able to consider the subject of VI comprehensively and to place them in the context
of future studies.
Limitations and future research
With regard to the chosen methodology, there are certain limitations. For this research, a
quantitative approach using an online laboratory experiment was chosen. The laboratory
experiment takes place in an artificial space, which allows the researchers to keep
variables of the experiment under control and to reduce potentially confounding factors.
This results in a high internal validity. However, it should be noted that the creation of an
artificial space limits the realism of the experiment. The laboratory experiment results in
a lower external validity than, for example, a field experiment.
In our experiment, only a relatively small sample could be examined. As a result,
the findings can only be applied to the general public to a limited extent. Likewise, only
a total of four influencers could be studied. Here, too, only limited general statements can
be made about reality. For a realistic representation, a renewed study with a higher
number of influencers and a larger number of test subjects would make sense.
The focus of this study has been on influencers from the fashion and lifestyle
sectors. Further research could be conducted here in greater depth by examining other
influencers with the same methodology and comparing them with the findings of this
work.
The research design of this paper investigated the extent to which SMI and VI are
perceived differently. However, the extent to which knowledge about the type of
influencer is a factor in the change in perception was not considered. Although VI were
shown in the course of the experiment, it was not communicated to the test persons that
the influencers shown were VI.
It would therefore be highly relevant to investigate in further studies to what extent
knowledge about the type of influencer has an impact on the evaluation of the test persons.
This insight could also be relevant for brand communication and the planning of
Copyright: Henry Stewart Publications 13
influencer marketing activities, in order to be able to make more informed decisions about
the appearance and presentation of the VI. For this purpose, further research in the design
of the experiment could, for example, include a third test group that looks at identical VI,
but is informed beforehand that they are virtual persons.
7
References
1.
Boyd, D. M., & Ellison, N. B. (2007). Social network sites: Definition, history,
and scholarship. Journal of computer‐mediated Communication, 13(1), 210-230.
https://doi.org/10.1111/j.1083-6101.2007.00393.x
2.
Rudeloff, C., Pakura, S., Eggers, F. et al. It takes two to tango: the interplay
between decision logics, communication strategies and social media engagement
in start-ups. Rev Manag Sci 16, 681–712 (2022). https://doi.org/10.1007/s11846-
021-00464-x
3.
Stefanie Pakura & Christian Rudeloff (2023) How entrepreneurs build brands and
reputation with social media PR: empirical insights from start-ups in Germany,
Journal of Small Business & Entrepreneurship, 35:2, 153-180, DOI:
10.1080/08276331.2020.1728490
4.
Pakura, S., Rudeloff, C., Bekmeier-Feuerhahn, S., & Eggers, F. (2020).
Communication management of start-up firms. An empirical analysis of
entrepreneurs' communication and networking success on Facebook. International
Journal of Entrepreneurial Venturing, 12(5), 459-489.
https://doi.org/10.1504/IJEV.2020.111538
5.
Ki, C.W.C. and Kim, Y.K. (2019), “The mechanism by which social media
influencers persuade consumers: the role of consumers’ desire to mimic”,
Psychology and Marketing, 36(10), 905-922, https://doi.org/10.1002/mar.21244
6.
Influencermarketinghub (2022). The State of Influencer Marketing 2022:
Benchmark Report. https://influencermarketinghub.com/influencer-marketing-
benchmark-report/
7.
Vrontis, D., Makrides, A., Christofi, M., & Thrassou, A. (2021). Social media
influencer marketing: A systematic review, integrative framework and future
research agenda. International Journal of Consumer Studies, 45(4), 617-644.
8.
Sands, S., Ferraro, C., Demsar, V., & Chandler, G. (2022). False idols: Unpacking
the opportunities and challenges of falsity in the context of virtual influencers.
Business Horizons.
9.
Travers, C. (2020). 5 Notable Virtual Influencers + Brand Partnerships.
https://www.virtualhumans.org/article/5-notable-virtual-influencer-brand-
partnerships
10.
Ahn, R. J., Cho, S. Y., & Sunny Tsai, W. (2022). Demystifying Computer-
Generated Imagery (CGI) Influencers: The Effect of Perceived
8
Anthropomorphism and Social Presence on Brand Outcomes. Journal of
Interactive Advertising, 1-9.
11.
Lazarsfeld, P. F., Berelson, B. R &, Gaudet, H. (1944). The People's Choice: How
the Voter makes up his Mind in a Presidential Campaign. Columbia University
Press. New York.
12.
Katz, E. & Lazarsfeld, P. F. (1955). Personal Influencer: The part played by the
people in the flow of mass communications. The Free Press. New York.
13.
Bennett, W. L. & Manheim, J. B. (2006). The One-Step Flow of Communication,
in The ANNALS of the American Academy of Political and Social Science (1st
ed. pp. 213-232). Vol. 608
14.
Leung, F.F., Gu, F.F. & Palmatier, R.W. Online influencer marketing. J. of the
Acad. Mark. Sci. 50, 226-251 (2022). https://doi.org/10.1007/s11747-021-00829-
4
15.
Park, J., Lee, J. M., Xiong, V. Y., Septianto, F., & Seo, Y. (2021). David and
Goliath: when and why micro-influencers are more persuasive than mega-
influencers. Journal of Advertising, 50(5), 584-602.
16.
Rundin, K., & Colliander, J. (2021). Multifaceted influencers: towards a new
typology for influencer roles in advertising. Journal of Advertising, 50(5), 548-
564.
17.
Yesiglou, S. & Costello, J. (2020). Influencer Marketing: Building Brand
Communities and Engagement. Routledge. New York.
18.
noonoouri [@noonoouri] (2022, June 24). noonoouri. [Instagram profile].
Instagram. Retrieved 2022-04-13 from https://www.instagram.com/noonoouri/
19.
Lou, C., & Yuan, S. (2019). Influencer marketing: how message value and
credibility affect consumer trust of branded content on social media. Journal of
Interactive Advertising, 19(1), 58-73.
20.
Chekima, B., Chekima, F. Z., & Adis, A. A. A. (2020). Social media influencers
in advertising: The role of attractiveness, expertise and trustworthiness. Journal
of Economics and Business, 3(4).
21.
Myers, S. (2021). Instagram Source Effects: The Impact of Familiarity and
Likeability on Influencer Outcomes. Journal of Marketing Development and
Competitiveness, 15(3), 50-55.
22.
Janssen, L., Schouten, A. P., & Croes, E. A. (2022). Influencer advertising on
Instagram: product-influencer fit and number of followers affect advertising
9
outcomes and influencer evaluations via credibility and identification.
International journal of advertising, 41(1), 101-127.
23.
Masuda, H., Han, S. H., & Lee, J. (2022). Impacts of influencer attributes on
purchase intentions in social media influencer marketing: Mediating roles of
characterizations. Technological Forecasting and Social Change, 174, 121246.
24.
Stein, J. P., Linda Breves, P., & Anders, N. (2022). Parasocial interactions with
real and virtual influencers: The role of perceived similarity and human-likeness.
New Media & Society, 14614448221102900.
Tables
Table I:
Overview of influencers
Influencer 1
Influencer 2
Influencer 3
Influencer 4
Name
Miquela Sousa
Blawko
Julia Marie
Marcel Floruss
Instagram Name
@lilmiquela
@blawko22
@xlaeta
@marcelfloruss
Follower
3.000.000
141.000
2.900.000
536.000
Topic
fashion/lifestyle
fashion/lifestyle
fashion/lifestyle
fashion/lifestyle
Category
mega influencer
mega influencer
mega influencer
mega influencer
Sex
Weiblich
Männlich
Weiblich
Männlich
Influencer Type
VI
VI
SMI
SMI
Note: Number of followers in September 2022
10
Table II
Group descriptives
Group
N
mean
SE
a
Credibility
VI
32
3.03
0.150
0.88
SMI
31
3.19
0.203
Expertise
VI
32
3.34
0.157
0.88
SMI
31
3.4
0.223
Likeability
VI
32
2.46
0.163
0.87
SMI
31
2.76
0.228
Purchase intentions
VI
32
2.55
0.194
SMI
31
2.6
0.242
... According to Kotler et al. (2016) human creativity and intuition cannot be fully replicated by AI, and therefore, in certain situations, human-managed campaigns may be more effective. Bohndel et al. (2023), along with Looi and Kahlor (2024), explore the role of virtual AI-driven influencers compared to traditional human influencers. While there are few differences in terms of credibility and competence, human influencers score higher in likability and engagement. ...
Article
Purpose- The rapid advancement in digital marketing, driven by technologies such as artificial intelligence (AI), forms the backdrop for this research. This study aims to investigate the performance differences between AI-driven and human-managed digital marketing campaigns by means of a true field experiment. Selected Key Performance Indicators (KPIs) are evaluated on the Meta platform to make a statement regarding the performance. Methodology- The study has an experimantal research method. Two concurrent marketing campaigns for the Paul Kenzie brand were conducted over a two-week period: one fully created by ChatGPT-4 and the other by a human expert. Key KPIs measured include Click-Through Rate (CTR), number of conversions, conversion rate, and Return on Advertising Spend (ROAS). Findings- The results indicate that AI-driven campaigns outperform human-managed campaigns in terms of CTR, conversion rate, and ROAS, suggesting higher efficiency and effectiveness in reaching and engaging the target audience. Conclusion- The findings highlight the potential of integrating AI technologies with human creativity to optimize digital marketing strategies. Keywords: Artificial ıntelligence, social media marketing, digital marketing, field experiment JEL Codes: M15, M31, Q55
... nevertheless, brands continued to leverage their influence to increase brand awareness and visibility on a larger scale. More recently, the concept of mega influencers has emerged, representing individuals with enormous followings and significant influence across multiple platforms (Böhndel et al., 2023). these influencers, often celebrities or public figures, command immense reach and influence, making them sought-after partners for brands aiming to make a big impact. ...
Article
Full-text available
Artificial intelligence represents the cutting-edge frontier in influencer marketing, offering a novel approach to leveraging social media data. This study aims to provide a comprehensive analysis of the recent state and future directions of artificial intelligence-infused influencer marketing through bibliometric analysis. For substantial bibliometric analysis, the study contains 316 documents published in journals indexed in the Scopus database for the period ranging from 2004 to 2023. The yearly publications have risen from 2 in 2004 to 55 in 2023, with an average age of 19.06 years and a citation rate of 22.19 yearly. The number of publications in this domain accelerated in 2020. According to the review findings, Hong Kong is the most productive country in this research domain. The leading affiliation is Hanyang University. The top contributing author is Gatica Perez-D. The most prolific source is the AAAI Spring Symposium Technical Report. Keyword occurrence analysis highlights prevalent terms such as Blogs, Social media networking (online) and Data mining alongside emerging keywords such as Influencer marketing, User-Generated Content, Artificial Intelligence, Sentiment Analysis, and Semantic Analysis in this domain to anticipate future inclinations. Performance analysis, further delves into Collaborative efforts among countries, sources, affiliations and authors have also been done using tools like Biblioshiny and Vos Viewer. The study offers novel contributions to the existing literature in terms of comprehensively providing evidence of the current practices of AI-integrated influencer marketing.
... Based on our results, it can be assumed that potentially negative perceptions of AI disclosure may be attenuated by endowing AI artefacts with human-like features that are capable of inducing brand anthropomorphism. This is also reflected in prior studies (Konya-Baumbach et al., 2023) and can be illustrated regarding the new phenomenon of human-like AI-generated social media influencers, which can exert positive brand benefits similar to those produced by human celebrity endorsers (Böhndel et al., 2023). These endorsers possess distinct human-like visual representations, thus enabling them to induce anthropomorphic perceptions. ...
Article
Full-text available
Technological innovations, including artificial intelligence (AI), are progressively being integrated into marketing and branding. However, the novel nature of these technologies often shrouds their potential benefits and associated risks. This holds also true for the implementation of AI in the domain of brand voices. Brand voices may play a crucial role in fostering human-like brand perception (brand anthropomorphism), potentially offering competitive advantages for companies. However, the integration of AI in shaping brand voices in an advertising context remains a relatively unexplored area. This study addresses this gap by examining the impact of pre-recorded and AI-generated synthetic brand voices on brand anthropomorphism and brand equity. Conducted through a 3 × 2 online experiment the study shows that both pre-recorded and synthetic brand voices positively influence brand anthropomorphism. Additionally, brand anthropomorphism emerges as a strong mediator between brand voices and brand equity. Notably, the positive effects persist even when the synthetic brand voice is disclosed as being AI-generated. Theoretical and practical implications of these findings are derived.
Article
Full-text available
Dynamiczny rozwój rynkowy oraz technologiczny wpływa na wiele dziedzin życia. W ich wyniku wzrasta m.in. liczba marek, które konkurują między sobą o zdobycie uwagi klienta. Konsument XXI wieku pragnie z kolei, aby proces zakupowy nie miał jedynie charakteru materialnego, a jako interaktywny generował emocje, doświadczenia, zaś sam produkt był spersonalizowany. Odpowiedzią na te potrzeby stają się strategie sprzedażowe oparte na tzw. tworzeniu doświadczeń, wśród których można wymienić zyskujący na popularności storytelling (który coraz częściej rozwijany jest w przestrzeni internetowej). Z drugiej strony także dynamicznie rozwijające się systemy AI zmieniają podejście sprzedażowe. W wyniku tego coraz bardziej popularne w nowoczesnych strategiach marketingowych staje się połączenie storytellingu i tzw. wirtualnego influencer marketingu. Więcej firm decyduje się zatem na zaproszenie do współpracy tzw. wirtualnych ludzi, znanych jako „wirtualni influencerzy”, którzy kreowani są, aby jak najbardziej przypominać żywe osoby. Wirtualni influencerzy, wykorzystując pozory autentyczności, zaangażowanie społecznościowe i wzajemne zaufanie jako podstawowe walory, mają wyjątkowy potencjał do opowiadania angażujących, spójnych historii, co idealnie wpisuje się w technikę storytellingu. W obliczu rozwoju technologicznego następuje zatem rozbudowanie sposobów, poprzez które marki budują relacje z klientami i tworzą opowieści, promując swoje produkty.
Article
Full-text available
This study investigates the evolving role of artificial intelligence (AI) in consumer decision-making, particularly in comparison to traditional human influencers. As consumer trust in social media influencers has declined, largely due to concerns about the financial motivations behind endorsements, AI tools such as ChatGPT have emerged as perceived neutral intermediaries. The research aims to understand whether AI systems can replace human influencers in shaping purchasing decisions and, if so, in which sectors. A mixed-methods approach was employed, involving a quantitative survey of 478 participants with prior experience using both AI tools and interacting with social media influencers, complemented by 15 semi-structured interviews. The results reveal that AI is favoured over human influencers in product categories where objectivity and precision are critical, such as electronics and sporting goods, while human influencers remain influential in emotionally driven sectors like fashion and beauty. These findings suggest that the future of marketing will show a reduced need for human social media influencers and may involve a hybrid model where AI systems dominate data-driven recommendations and human influencers continue to foster emotional engagement. This shift has important implications for brands as they adapt to changing consumer trust dynamics.
Article
Full-text available
The advent of artificial intelligence (AI) technologies has profoundly impacted social media, particularly through the creation and deployment of AI-generated social media accounts. These accounts are designed to simulate human behavior and interaction, and their influence on social dynamics is increasingly significant. This paper provides a comprehensive review of how AI-generated social media accounts shape social influence by examining their technological foundations, practical applications, and associated ethical concerns. By exploring the functionality of AI-driven bots and deepfake profiles, we analyze their roles in political campaigning, consumer marketing, and the spread of misinformation. Additionally, the paper discusses the ethical implications and proposes regulatory measures to address the challenges posed by these technologies.
Article
Full-text available
While the popularity of virtual influencers in influencer marketing is on the rise, little is known about the reasons behind their increasing popularity. Furthermore, there is also a dearth of research examining why and how virtual influencers could be more effective than human influencers in influencing consumer purchase decisions. Drawing on the concept of perceived authenticity of influencers and machine heuristic, this study investigates the effects of influencer type (virtual influencer vs. human influencer) on the perceived authenticity of consumers. In addition, we also explore if machine heuristic will moderate the perceived authenticity of consumers and whether trust in influ-encers and purchase intentions will be subsequently associated with the perceived authenticity of the influencer. Results from an online between-subjects designed experiment (N ¼ 130) indicated that virtual influencers were unexpectedly perceived as more authentic than human influencers. Additionally, the perceived authenticity of virtual influencers significantly increased among those with a strong belief in the machine heuristic. Subsequently, we found a positive association between the perceived authenticity of influencers and the extent of consumers' trust in them. Additionally, trust in influencers was positively associated with consumers' intention to purchase the products they promoted by the influencers. These findings provide valuable insights into the use of virtual influencers for marketing purposes. Further theoretical and practical implications are discussed.
Article
Full-text available
Research confirms that over 70% of American adults actively engage on social media (Pew Research, 2019). As social media use grows, marketers continue to invest heavily in digital strategies including influencer marketing (Balilis, 2020). These influencers have created a new economy where brands pay to tap into the connections they have with followers. This study extends the source attractiveness model to top Instagram influencers using data from You.gov to measure familiarity and likeability of a sample of 82 top Instagram influencers with between 20 and 200 million followers. Findings suggest that these source characteristics impact the number of followers and engagement with content. Specifically, the results indicate that more familiar influencers had more followers but lower engagement while more likeable influencers had fewer followers but higher engagement. This indicates that familiarity and likeability may lead to different social media outcomes.
Article
Full-text available
Influencer marketing has become a dominant and targeted means for brands to connect with consumers, but it also brings risks associated with influencer transgression and reputation damage. In recent years, virtual influencers have gained popularity and given rise to falsity, or artificially created and manipulated influencers that are revolutionizing the field of influencer marketing. A virtual influencer is an entity—humanlike or not—that is autonomously controlled by artificial intelligence and visually presented as an interactive, real-time rendered being in a digital environment. As brands increasingly seek to engage virtual influencers to connect with and sell to audiences, we take a step back and discuss the opportunities and challenges they present for firms and managers. To help marketers understand this emerging field, we first document the rise of virtual influencers. Next, we discuss consumer reactions to virtual influencers before unpacking their associated opportunities and challenges for brands and marketers. Finally, we conclude with an overview of implications and future considerations.
Article
Full-text available
Digitally created online celebrities (so-called virtual influencers) have appeared on various social media and video streaming platforms. While the scientific community has recently started to take an interest in this new phenomenon, it still remains mostly unclear how online audiences engage with—and relate to—these artificial digital creations. To address the identified research gap, we conducted a preregistered experiment (N = 179), comparing viewers’ parasocial interactions (PSIs) with either a human or a virtual influencer. Based on natural stimuli, we find that viewers’ parasocial response does not differ significantly between the two groups. However, by focusing on several theoretically relevant mediator variables, we uncover two opposing effects at play: While a significant direct effect signifies stronger PSIs with the virtual influencer, participants also attribute this persona with less mental human-likeness and similarity to themselves—which ultimately suppresses the observed advantage. Potential explanations for our results are discussed.
Article
Full-text available
Social media influencers are increasingly employed as product endorsers, and a growing body of academic research confirms that influencers are an effective advertising instrument. However, more research is needed on the specific influencer characteristics driving this success, and the processes responsible for these effects. In this study, we investigated to what extent product-influencer fit and number of followers interact in contributing to positive advertising outcomes and influencer evaluations. Moreover, we investigated to what extent perceived credibility of the influencer and identification with the influencer mediate these relationships. We conducted a 2 (poor vs. good product-influencer fit) X 2 (moderate vs. high number of followers) between-subjects experiment among 432 Dutch Instagram users. Participants were exposed to Instagram posts of health and fitness influencers who endorsed either a protein shake (good-fit) or ice cream (poor-fit). Results showed that fit and number of followers seem to work in tandem: although influencers with a high number of followers are liked more, and their endorsements result in a more positive attitude toward both the ad and product, and a greater likelihood to buy the advertised product as compared to influencers with a moderate number of followers, the endorsed product should fit the influencer’s self-branded image for these positive effects to occur. Moreover, effects were all mediated by perceived credibility and identification, which appear to be important drivers of influencer endorsement effects. Together, our findings contribute to a growing body of knowledge on the processes driving audience responses to influencer marketing, and provide clear guidelines for practitioners.
Article
Full-text available
This manuscript analyzes start-ups’ usage of different communication strategies (information, response, involvement), their underlying decision logics (effectuation, causation, strategy absence) and respective social media success. A multitude of studies have been published on the decision logics of entrepreneurs as well as on different communication strategies. Decision logics and according strategies and actions are closely connected. Still, research on the interplay between the two areas is largely missing. This applies in particular to the effect of different decision logics and communication models on social media success. Through a combination of case studies with fuzzy-set Qualitative Comparative Analysis this exploratory study demonstrates that different combinations of causal and absence of strategy decision logics can be equally successful when it comes to social media engagement, whereas effectuation is detrimental for success. Furthermore, we find that two-way-communication is essential to create engagement, while information strategy alone cannot lead to social media success. This study provides new insights into the role of decision logics and connects effectuation theory with the communication literature, a field that has been dominated by causal approaches.
Article
Computer-generated imagery (CGI) influencers are at the forefront of brand communication trends driven by artificial intelligence (AI). Focusing on Lil Miquela, this study explores the mechanism underlying the CGI influencer’s persuasiveness based on the computers-are-social-actors (CASA) framework. The study examines how consumers’ attitudes toward the influencer’s sponsored posts and the endorsed brands are impacted by her perceived anthropomorphism through mediators of social presence and perceived attractiveness. Based on a survey of generation Z consumers, we found that the CGI influencer’s perceived anthropomorphism effectively enhances her social presence, which in turn boosts her perceived physical and social attractiveness, respectively, to drive consumer evaluation outcomes.
Article
Online influencer marketing (OIM) has become an integral component of brands’ marketing strategies; however, marketers lack an adequate understanding of its scope, effectiveness, and potential threats. To fill this gap, this article first describes the phenomenon’s background, defines OIM, and delineates its unique features to set the conceptual boundaries for the new concept. Drawing insights from practitioner and consumer interviews, and in line with social capital theory, the authors propose that OIM can be understood as leveraging influencer resources (including follower networks, personal positioning, communication content, and follower trust) to enhance a firm’s marketing communication effectiveness. Six novel propositions illustrate the benefits and potential threats of OIM, which may spur progress toward a theory of OIM. Finally, this article outlines key strategies for effectively managing OIM and identifies important literature–practice gaps to suggest avenues for further research.
Article
Social media influencers (SMIs) have become an efficient advertising tool. However, their roles vis-à-vis the brands for which they advertise are changing. Far from being just simple promoters anymore they now take active part in both the product and communications development of firms. Yet in advertising research, a discussion of these new roles is absent. This article, therefore, seeks to categorize the roles influencers play in advertising collaborations by means of an empirical typology. In a netnographic study, we investigate the posts of 21 fashion influencers over a period of three years and find three main roles (spokesperson, cocreator, or co-owner) and eight subroles (for spokesperson: billboard, stylist, and ambassador; for cocreator: consultant and codesigner; for co-owner: sole proprietor, facilitator, and partner) that influencers take on in relation to brands. We contribute to theory on influencer marketing by conceptualizing influencer roles in advertising collaborations as well as categorizing these roles, thus allowing future researchers to use our typology as a conceptual foundation. Another contribution is the insight into how the dynamics of control over product and communication creation is evolving in influencer collaborations.
Article
Social media influencer (SMI) advertising is on the rise; however, extant theory regarding the determinants of SMI advertising effectiveness is undeveloped. The present research establishes when and how the type of SMI based on the number of followers influences SMI advertising effectiveness. Specifically, the findings of four experimental studies show that micro-influencers (those who have 10,000 to 100,000 followers) are more persuasive than mega-influencers (those who have more than 1 million followers) because endorsements by micro-influencers (versus mega-influencers) bestow higher perceptions of authenticity on the endorsed brand, which “rubs off” from the perceptions regarding influencer authenticity. However, this differential effect of SMI type through influencer and brand authenticity occurs only when the endorsed product is perceived as hedonic (as opposed to utilitarian) consumption. From a theoretical perspective, these findings extend prior research on how endorsements from different SMI types vary in terms of their persuasiveness, shed light on the underlying mechanism, and identify consumption type as an important boundary condition. From a practical perspective, we offer managerial implications for enhancing the effectiveness of SMI advertising strategies by taking into consideration SMI types, consumption contexts, and message framing styles.
Article
Social media influencer marketing has recently received significant attention. Many studies have explored the parasocial relationship (PSR) formation between influencers and followers. PSR has not often been weighted against other widely used relationship marketing constructs, despite the multitude of PSR studies. This study developed a research model based on the theory of persuasion, which was constructed to investigate the relative weight of the PSR. The study considered three personal attributes (attitude homophily, physical attractiveness, and social attractiveness) and three characterizations (trustworthiness, perceived expertise, and PSR) as antecedents of purchase intention. Data were collected through a survey of respondents who bought products/services after watching YouTube advertisements made by influencers. The study found that PSR had a significantly positive impact on purchase intentions relative to other characterizations and that PSR was significantly related to the three personal attributes. In addition, PSR formation was significantly influenced by consumers’ perceived influencer types. The survey showed that social media influencer marketing strategies need to be fine-tuned based on personal attributes, characterizations, and influencer types. This paper discusses the theoretical and practical implications of these findings.