The effectiveness of virtual versus human influencer marketing for tourism destinations

Abstract

The organisations that manage tourism destinations are increasingly working with influencers in their promotional campaigns. Our research analyses how effective the promotion of a destination is according to the type of influencer: human or virtual. We used a design of experiments approach in which 954 participants from generations Y and Z looked at posts of tourism destinations created by the two types of influencers on Instagram. The results suggest that source credibility, parasocial interaction, attitude towards the post and attitude towards the destination are antecedents, direct or indirect, of intention to travel to the destination and to recommend it. The results show that both types of influencers are effective at promoting the destinations, although it was also confirmed that a human influencer is more effective at promoting a destination than a virtual one. Therefore, this study has practical implications for tourism organisations.

Keywords

Virtual influencer promotion tourism destination source credibility parasocial interaction

Introduction

Among the promotional campaigns undertaken by the organisations that manage tourism destinations, we can highlight collaboration with influencers who promote the destination via social networks (Gretzel, 2018; Yılmaz et al., 2020). Through influencer marketing, organisations can communicate with and spread their message to the target public directly and in a more relatable way (Childers et al., 2019). Recently, the emerging development of artificial intelligence in the online environment has enabled the creation of virtual influencers (Appel et al., 2020) who look and behave like humans (Arsenyan and Mirowska, 2021).

In the context of tourism, the application of artificial intelligence (AI) has become a trending topic and one of interest in the academic and business spheres. On the one hand, from the tourist's point of view, AI can improve their experience by facilitating the decision-making process (Bulchand-Gidumal, 2022). On the other hand, companies can use AI in relevant areas such as advertising (Bulchand-Gidumal, 2022; Franke et al., 2023). The organisations that manage tourism destinations can also benefit from applying this technology to their communication campaigns (Bulchand-Gidumal, 2022; Tussyadiah, 2020). More specifically, advances in AI have made it possible to create virtual influencers. These influencers showcase a similar lifestyle to that of human influencers and interact with followers in a similar way (Arsenyan and Mirowska, 2021). However, unlike the promotional activities carried out by human influencers, the messages conveyed by virtual influencers are managed entirely by the company in question, which enables greater control over brand reputation (Jhawar et al., 2023). Recently, brands like Prada and Calvin Klein have chosen virtual influencers like Lil Miquela (@lilmiquela) to promote their products on social networks (Drenten and Brooks, 2020). In our field of interest, Italy has promoted its tourism via a virtual influencer called Venere (@venereitalia23) who has 243,000 followers. However, because this field of study into AI is new, research into virtual influencers (Lou et al., 2022; Sands et al., 2022a) and their comparison with human influencers is very scarce (Arsenyan and Mirowska, 2021; da Silva Oliveira and Chimenti, 2021; Franke et al., 2023; Thomas and Fowler, 2021).

In the area of tourism, where there are numerous options for developing AI (Tussyadiah, 2020), no studies have been conducted on virtual influencers, analysing the effectiveness of the messages that they convey and their effect on decision-making in terms of visiting and recommending a destination. Our research intends to address the gap in this area of study by analysing how effective the promotion of a destination is according to the type of influencer: human or virtual. This model includes the theories of source credibility and parasocial interaction (PSI) and applies them in a novel way to the study of virtual influencers in the tourism sector. First of all, we propose a theoretical model applied to both types of influencers in which we intend to find out how different antecedent variables (source credibility, PSI, attitude towards the post and attitude towards the destination) influence intention to travel and recommend a destination. Secondly, based on a research question, we intend to study whether there are differences in the variables analysed depending on the type of influencer. To corroborate the hypotheses proposed in the theoretical model and address the research question posed, we used a design of experiments approach which involved answering a questionnaire after looking at posts by two influencers (one human and one virtual) showing the same tourism destination on Instagram.

Literature review

Virtual influencers

In recent years, academic and business interest in virtual influencers (VIs) has increased. According to Thomas and Fowler (2021), we can define an ‘AI influencer as a digitally created artificial human who is associated with Internet fame and uses software and algorithms to perform tasks like humans’ (p. 2). Likewise, Sands et al. (2022a) point out that ‘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’ (p. 778). Like human influencers, the activity of VIs aims to promote the brand on social networks, directly influencing the behaviour of their followers (Leung et al., 2022) and benefiting the brand (Thomas and Fowler, 2021). These influencers, designed to look and behave like humans (Franke et al., 2023; Lou et al., 2023), can also give companies a greater sense of security as using them enables them to have full control over their communication.

One of the most well-known examples of a VI who has collaborated with different brands is Lil Miquela, created in 2016 (Sands et al., 2022a). This influencer, who has 2.6 million followers on Instagram, defines herself as a ‘19-year-old Robot living in LA’ (Miquela, @lilmiquela) and in 2018 she was recognised as a relevant figure on the Internet (Thomas and Fowler, 2021; Time, 2018).

In the scarce research conducted in this field of study, some authors have focused on analysing the differences between virtual and human influencers. For example, Arsenyan and Mirowska (2021) compared three types of influencers: anime-like VIs; human-like VIs; and human influencers. The results of the study reveal that fewer positive reactions are received by the VI with a human appearance than by the human influencer. Sands et al. (2022b) aimed to analyse consumers’ perceptions of AI influencers and human influencers. The authors did not find any differences in two variables of interest in influencer marketing: intention to follow and personalisation. On the other hand, they observed differences in the variables source trust and WOM, trust being greater in the case of the human influencer and WOM greater in that of the AI influencer. For their part, Seymour et al. (2020) compared VIs with human ones via an experimental study analysing three specific variables: trustworthiness, affinity and preference. The participants displayed more trust, affinity and preference in the case of the human influencer. Finally, Franke et al. (2023) compared virtual and human influencers in terms of their attractiveness. The results revealed that users showed a more favourable attitude towards publicity campaigns fronted by human influencers.

Research hypotheses

Credibility

Previous studies consider credibility to be a construct composed of different dimensions: attractiveness, expertise, trustworthiness and similarity (Munnukka et al., 2016; Ohanian, 1990; Yılmazdoğan et al., 2021; Yuan and Lou, 2020). The first, expertise, can be defined as ‘knowledge and experience the person has in the given domain’ (p. 2) (Sokolova and Kefi, 2020). The second, trustworthiness, is related to the honesty and trust perceived by the audience in terms of the source (Teng et al., 2014). The third, the attractiveness of the source, may play a relevant role in the acceptance of the message (Teng et al., 2014). The last, similarity, is defined as the perception of similarities between the source and the recipient (Munnukka et al., 2016). Several authors have analysed the existing relationship between credibility and the intention to behave a certain way (Sokolova and Kefi, 2020; Weismueller et al., 2020). For example, in a study on social networks, Sokolova and Kefi (2020) demonstrated a positive relationship between credibility and purchase intention. Moreover, the results of the study on influencers by Weismueller et al. (2020) confirm the relationship between purchase intention and attractiveness, trustworthiness and expertise. More specifically, in the context of tourism, in a research paper on influencers, Yılmazdoğan et al. (2021) confirmed a significant relationship between trustworthiness and expertise in terms of intention to travel. Therefore, we propose the following research hypothesis:

H1a: The influencer's credibility is positively related to the intention to travel.

Other authors have considered the variable PSI in their studies on influencers (Jin et al., 2021; Kim, 2022; Lin et al., 2021). Horton and Richard Wohl (1956) define this variable as the ‘illusion of face-to-face relationship with a media personality’ (p. 215). In the context of influencer marketing, PSI refers to the interaction between the follower and the influencer (Sokolova and Kefi, 2020). Several authors have found a significant relationship between credibility and PSI. For example, the results of the study by Yılmazdoğan et al. (2021) confirm that the trustworthiness and expertise dimensions have an influence on PSI. For their part, Lee and Watkins (2016) demonstrated a link between physical attractiveness and PSI. Based on these results, we propose the following research hypothesis:

H1b: The influencer's credibility is positively related to PSI.

Attitude – an individual's positive or negative assessment – has been considered to be a relevant variable when explaining behaviour (Ajzen, 2001). In influencer marketing, studying attitude towards the advertisement is key (Gong and Li, 2017; Lee and Kim, 2020). Previous studies have indicated that the different constructs that make up credibility – attractiveness, expertise and trustworthiness – can affect acceptance of the message (Teng et al., 2014). Furthermore, in the context of tourism, a study by Shang and Luo (2021) confirmed that the influencer's credibility and attitude towards the destination are positively related. As indicated by previous studies, we propose the following two research hypotheses:

H1c: The influencer's credibility is positively related to attitude towards the post.

H1d: The influencer's credibility is positively related to attitude towards the destination.

Parasocial interaction

One of the variables addressed in the study of influencers is PSI (Bhattacharya, 2023; Jin and Ryu, 2020; Kim, 2022). PSI, which occurs when there is a temporary emotional connection between the celebrity and their followers (Penttinen et al., 2022; Zhang et al., 2020), has been positively related to attitude towards the post (Gong and Li, 2017). Moreover, previous research has shown that there is a positive relationship between PSI and purchase intention (Jin and Ryu, 2020; Kim, 2022; Lin et al., 2021; Sokolova and Kefi, 2020). Specifically, in the context of tourism, Yılmazdoğan et al. (2021) show that PSI has a significant influence on intention to travel. Therefore, in this study, we have formed the following research hypotheses:

H1e: PSI is positively related to attitude towards the post.

H1f: PSI is positively related to the intention to travel.

Attitude towards the post and the destination

In the research into the attitude towards a brand/product, attitude towards advertising plays a fundamental role (Munnukka et al., 2016). In turn, attitude towards the brand/product influences the intention to purchase and recommend it (Belanche et al., 2021). In tourism, a number of authors have demonstrated a positive relationship between the variable's attitude towards the destination and intention to travel, as well as between attitude towards the destination and intention to recommend it (Alipour et al., 2020; Jalilvand et al., 2012). Based on this previous literature, we propose the following hypotheses:

H1g: Attitude towards the post is positively related to attitude towards the destination.

H1h: Attitude towards the destination is positively related to intention to travel.

H1i: Attitude towards the destination is positively related to the intention to recommend.

Theoretical model

In accordance with the literature review, Figure 1 shows the proposed conceptual model which presents the research hypotheses set out above.

Research question

Based on the proposed theoretical model (Figure 1), our initial intention was to detect the differences in the variables analysed according to the type of influencer. Given the scarcity of previous studies comparing the two types of influencers (Arsenyan and Mirowska, 2021; Franke et al., 2023; Seymour et al., 2020), we propose the following research question:

RQ1. Are there any differences in the variables analysed according to the type of influencer? More specifically, compared to a human influencer, will a VI have lower scores for the following variables: credibility (RQ1a), PSI (RQ1b), attitude towards the post (RQ1c), attitude towards the destination (RQ1d), intention to travel to the destination they are promoting (RQ1e) and intention to recommend it (RQ1f)?

Methodology

Design of experiments

Data was gathered using self-administered online questionnaires addressed to people in generations Y and Z between November 2022 and January 2023. The data was obtained using convenience and snowball sampling. In this study, we used a design of experiments approach. The experiment consisted of looking at Instagram posts by two influencers (one human and one virtual) showing an image of the same destination. These posts were created by a designer based on the Instagram profile of Lil Miquela (https://www.instagram.com/lilmiquela), in the case of the VI, and using an image bank in that of the human influencer. The VI had human features but was identified as a robot in their profile details. To avoid possible bias due to the type of destination, half of the sample looked at an urban destination and the other half a rural destination. The order of viewing was also alternated, showing the human influencer's post to half of the sample first and the VI's post to the other half and vice versa. After each image, a series of questions were asked concerning the constructs analysed (credibility, PSI, attitude towards the post, attitude towards the destination, and intention to travel and to recommend the destination). The posts by the two influencers were also edited to control the effect of other variables: number of posts, followers, people followed and the image of the destination. This means that the only relevant difference between one post and the other was the influencer's image (see Figure 2).

A pre-test was performed on 22 volunteers to detect any misunderstandings or ambiguity (Elmashhara and Soares, 2022). To be able to complete the questionnaire, the participants had to be active users of Instagram, follow influencers and have travelled outside of their place of residence for leisure during the last year.

Measurement

To ensure that the content was valid, different scales validated in previous studies were adapted for the purposes of our research. Table 1 contains the items used for each construct and the bibliographical sources used. These variables were measured using seven-point Likert-type scales, whereby 1 = “Strongly disagree” and 7 = “Strongly agree”.

Table 1.

Measurement scale.

Variable		Item	Reference
Attitude towards the post ATTP	ATTP1	I think the post is very good	Adapted from Munnukka et al. (2016) and Mitchell and Olson (1981)
	ATTP2	I think the post is pleasant
	ATTP3	I think the post is interesting
Intention to recommend REC	REC1	I will recommend this place shown by the influencer to others	Adapted from Belanche et al. (2021) and Bigne et al. (2001)
	REC2	I will say positive things to others about the place shown by this influencer
	REC3	I will encourage my friends and family to visit this place
Intention to travel INT	INT1	If I had the chance to travel, I would visit the place shown by this influencer	Adapted from Chen et al. (2014)
	INT2	When I travel, it's highly likely that I’ll visit the place shown by this influencer	Adapted from Chen et al. (2014)
	INT3	In future, I intend to book a trip to the place shown by this influencer	Authors’ own creation
Parasocial interaction PAR	PAR1	I look forward to seeing this influencer's videos, posts and comments	Adapted from Sokolova and Kefi (2020)
	PAR2	I would follow and interact with this influencer on other social networks	Adapted from Sokolova & Sokolova and Kefi (2020)
	PAR3	I would love to interact frequently with this influencer	Authors’ own creation
Credibility	CON1	I think that the influencer is honest and can be trusted	Adapted from Munnukka et al. (2016) and Ohanian (1990)
	EXP1	I feel that the influencer knows a lot about the destination they are showing	Adapted from Munnukka et al. (2016) and Ohanian (1990)
	ATR1	I think that the influencer is very attractive	Adapted from Munnukka et al. (2016) and Ohanian (1990)
	SIM1	The influencer and I have a lot in common	Adapted from Munnukka et al. (2016) and Bower and Landreth (2001)

Sample

The final sample consisted of 954 responses. Table 2 shows the profile of the sample. In terms of gender, 40.0% of the sample were male and 60.0% of the sample were female. According to age range, 58.0% of the participants were from 18 to 24 years of age, that is, from generation Z, and 42.0% were from 25 to 40 and, therefore, from generation Y. As far as their occupation was concerned, 6.9% were entrepreneurs or self-employed, 5.1% were managers or supervisors in companies, 26.4% were employed by a third party, 56.2% were students and 5.3% were unemployed. The socioeconomic status of 9.4% was below average, that of 74.3% average and that of 16.2% above average.

Table 2.

Sample structure.

Sex
Female	60.0%
Male	40.0%
Age
18–24 Gen. Z	58.0%
25–40 Gen. Y	42.0%
Education
Primary/secondary	5.6%
A levels/vocational training	28.1%
University	66.4%
Occupation
Entrepreneur/self-employed	6.9%
Manager/supervisor	5.1%
Employee	26.4%
Student	56.2%
Unemployed	5.3%
Economic status
Below average	9.4%
Average	74.3%
Above average	16.2%
Total	954

To verify the suitability of the sample size, we used G*Power (Faul et al., 2009). The results show that the size of the sample used (954) is sufficient to study the proposed model (six predictors) given that it exceeds the proposed sample size of 146 individuals which is required to obtain a statistical power of 0.95. Therefore, we can safely conclude that the sample size used (954) is far greater than that required for the purposes of this study.

Data analysis

First of all, to confirm the research question RQ1, we performed a Student's t test to check the differences in the means of the construct items, as well as Levene's test for equality of variances.

Secondly, we applied the partial least squares technique (PLS-SEM) using the software Smart PLS v.4.0.9.6 (Ringle et al., 2022) to evaluate the proposed theoretical model and research hypotheses. We have proposed a first-order model in which reflective constructs (mode A) and one formative construct (mode B) coexist, which supports the use of the PLS technique for their analysis (Sarstedt et al., 2016; Chin, 2010). We also analysed the global model fit taking the standardised root mean squared residual (SRMR) values as an approximate measure of model fit for PLS-SEM (Henseler et al., 2016), the VIF values (variance inflation factor) to check that there was no multicollinearity between the constructs, the measurement model by means of the reliability and validity of the reflective constructs, and the weighting and VIF of the formative constructs. In addition, the structural model was analysed by means of the R², the path coefficients and the confidence intervals. We performed this analysis for each type of influencer studied.

Results

Mean difference according to influencer type (human vs. virtual)

The data obtained from the descriptive analysis (mean) of the construct items of the proposed model for each of the influencers analysed (human and virtual) are shown in Table 3. The table also presents the results of the Student's t test conducted to check the differences in the means, taking into account different variances (negative Levene's test for equality of variances). The values of all the construct items considered are higher and statistically significant in the case of the human influencer.

Table 3.

Descriptive analysis, differences in the means and results of the measurement model assessment.

		Mean
		Human	Virtual	Dif.	Sig.	Loading	Composite reliability	AVE
Reflective Mode A constructs
ATTP	Attitude towards the post						0.897	0.743
ATTP1	I think that the post is very good	4.21	3.60	0.61	***	0.869
ATTP2	I think that the post is nice	5.05	4.36	0.69	***	0.850
ATTP3	I think that the post is interesting	4.09	3.72	0.37	***	0.867
PAR	Parasocial interaction						0.920	0.793
PAR1	I’m looking forward to seeing this influencer's videos, posts and comments	3.04	2.63	0.41	***	0.862
PAR2	I would follow and interact with her on other social networks	2.79	2.30	0.49	***	0.907
PAR3	I would love to interact frequently with this influencer	2.59	2.15	0.43	***	0.901
ATTD	Attitude towards the destination						0.898	0.746
ATTD1	The post gives me a positive impression of the place shown by this influencer	4.97	4.29	0.68	***	0.827
ATTD2	Based on the post, I think I would enjoy the place she is showing	4.85	4.43	0.42	***	0.880
ATTD3	Based on the post, I like the place this influencer is showing	4.92	4.56	0.36	***	0.884
INT	Intention to travel						0.912	0.775
INT1	If I had the chance to travel, I’d visit the place shown by this influencer	4.51	4.14	0.37	***	0.882
INT2	When I travel, it's highly likely that I’ll visit the place shown by this influencer	3.93	3.60	0.33	***	0.890
INT3	In future I intend to book a trip to the place shown by this influencer	3.77	3.47	0.30	***	0.869
REC	Intention to recommend						0.918	0.789
REC1	I will recommend this place shown by the influencer to others	3.89	3.41	0.49	***	0.872
REC2	I will say positive things to others about the place shown by this influencer	4.03	3.73	0.30	***	0.895
REC3	I will encourage my friends and family to visit this place	3.83	3.54	0.29	***	0.897
Formative Mode B construct						Weights	Sig.
CRE	Credibility
CON1	I think that the influencer is honest and can be trusted	3.79	3.01	0.79	***	0.293	***
EXP1	I feel that the influencer knows a lot about the destination she is showing	3.22	2.71	0.52	***	0.353	***
ATR1	I think that the influencer is very attractive	4.74	3.06	1.68	***	0.353	***
SIM1	The influencer and I have a lot in common	3.03	2.32	0.71	***	0.302	***

Scale from 1 to 7.

AVE: average variance extracted.

Significance level: *** p < 0.01; ** p < 0.05; * p < 0.1; ns: non-significant.

Attitude towards the destination and attitude towards the post are the constructs whose items obtained the highest values, with values above the middle of the scale (4); conversely, PSI and credibility present lower values, below the middle of the scale. As far as the constructs analysed are concerned, credibility presents the greatest differences between the human and the VI.

Table 4 shows the distribution of the frequencies of responses to the questions comparing the variables in relation to the human and the VI. We can see that, for all the variables, the human influencer has considerable advantages over the VI. Participants perceived the human influencer to be more trustworthy (difference of 70.3%), have more experience of the destination (48.2%), be more attractive (68.7%) and have more things in common with them (60.1%). Furthermore, those surveyed express a greater desire to interact with her (55.3%) and have a more positive attitude towards her post (50.7%) and to the place that she is showing (48.7%). Therefore, they are more willing to visit the place (54.6%) and recommend it (58.6%).

Table 4.

Comparison of virtual influencer and human influencer.

		Virtual more	The same	Human more	Dif.
Credibility	She inspires more trust	4.1%	21.5%	74.4%	70.3%
	She has more experience of the destination	7.5%	36.7%	55.8%	48.2%
	She's the most attractive influencer	5.9%	19.6%	74.5%	68.7%
	She has more things in common with me	3.8%	32.4%	63.8%	60.1%
Parasocial interaction	I have a stronger desire to interact with her	6.4%	31.9%	61.7%	55.3%
Attitude towards the post	I like her publication more	8.7%	31.9%	59.4%	50.7%
Attitude towards the destination	The place she is showing gives me a good impression	5.0%	41.2%	53.8%	48.7%
Intention to travel	She makes me want to travel to the place even more	4.3%	36.8%	58.9%	54.6%
Intention to recommend	She makes me more likely to recommend the place	3.8%	33.9%	62.4%	58.6%

Thus, the differences in the variables analysed according to the type of influencer responded to the research question (RQ1). Compared to the human influencer, the VI has lower scores in terms of credibility (RQ1a), PSI (RQ1b), attitude towards the post (RQ1c), attitude towards the destination (RQ1d), intention to travel (RQ1e) and intention to recommend the destination (RQ1f).

Relationships in the structural model

We performed an assessment of the global model, first of all, followed by the measurement model and, finally, the structural model. These analyses enabled us to assess the proposed structural model as well as the different research hypotheses.

Global model assessment

To assess the global model, we considered the SRMR, common method bias (CMB) and VIF (variance inflation factor) scores.

The SRMR scores were calculated as an approximation of the model fit in PLS-SEM (Henseler et al., 2016). The results indicate an acceptable fit (Henseler et al., 2016) as the final SRMR score is below 0.08, which is the recommended maximum (0.051).

We then used Harman's single-factor test (Podsakoff and Organ, 1986) to analyse the CMB. CMB is present if a single or general factor appears to represent most of the variance. A no-rotation factor analysis with the criterion of an eigenvalue greater than one revealed three different factors that represented 49.4%, 8.8% and 7.1% of the variance, respectively.

Finally, as the VIF scores were below 5, we were able to confirm that there were no issues of multicollinearity between the antecedent variables of each of the endogenous constructs.

Measurement model assessment

First of all, we analysed the reflective Mode A constructs and saw that the measurement model met the reliability and validity requirements recommended in previous literature (Chin, 2010; Hair et al., 2011, 2019; Gold et al., 2001): loadings (> 0.70), composite reliability (> 0.70), convergent validity (> 0.5) and heterotrait-monotrait (HTMT < 0.90).

We then analysed the formative Mode B construct ‘credibility’. In this case, we did not see any issues of multicollinearity between the indicators as the VIF of the construct items was lower than 3.3 (Diamantopoulos and Siguaw, 2006).

Table 3 shows that all the indicators make a significant and relevant contribution to the formation of the construct. The influencer's experience is the indicator that contributes the most (0.374), followed by similarity (0.337), attractiveness (0.316) and trust (0.271). Therefore, the reliability and convergent and discriminant validity requirements are met in the measurement model.

Structural model assessment

We started by using the bootstrapping technique to assess the relationships in the structural model and analyse the confidence intervals (Henseler et al., 2009). We used 10,000 samples and conducted a one-tailed test (Hair et al., 2011).

All the proposed relationships are significant (see Table 5 and Figure 3). The following have a positive influence on intention to travel: credibility (H1a: β = 0.152, p < 0.001, f² = 0.019); PSI (H1f: β = 0.169, p < 0.001, f² = 0.029); and attitude towards the destination (H1h: β = 0.462, p < 0.001, f² = 0.244). A significant, positive relationship is observed between attitude towards the destination and intention to recommend (H1i: β = 0.643, p < 0.001, f² = 0.706) and between credibility and PSI (H1b: β = 0.670, p < 0.001, f² = 0.812). Credibility (H1d: β = 0.175, p < 0.001, f² = 0.041) and attitude towards the post (H1e: β = 0.640, p < 0.001, f² = 0.544) have an influence on attitude towards the destination. Finally, credibility (H1c: β = 0.507, p < 0.001, f² = 0.279) and PSI (H1g: β = 0.253, p < 0.001, f² = 0.069) have an influence on attitude towards the post. Therefore, the results support all the proposed research hypotheses.

Table 5.

Hypothesis test, variance decomposition, Q² redundancy and effect size results.

Hypothesis	Relationships	Path coefficient	Sig.	T statistics	Confidence intervals	Confidence interval bias	Supported	Variable correlation	R ²	Q ²	f ²
Intention to travel									0.460	0.297
H1a	Credibility → Intention to travel	0.152	***	5.547	[0.106; 0.198]	[0.104; 0.195]	Yes/Yes	0.546	0.083		0.019
H1f	Parasocial interaction → Intention to travel	0.169	***	7.145	[0.13; 0.208]	[0.131; 0.209]	Yes/Yes	0.495	0.084		0.029
H1h	Attitude towards the destination → Intention to travel	0.462	***	20.102	[0.424; 0.499]	[0.423; 0.498]	Yes/Yes	0.636	0.294		0.244
Intention to recommend									0.413	0.307
H1i	Attitude towards the destination → Intention to recommend	0.643	***	40.854	[0.617; 0.67]	[0.616; 0.668]	Yes/Yes	0.643	0.413		0.706
Parasocial interaction									0.449	0.448
H1b	Credibility → Parasocial interaction	0.670	***	50.181	[0.648; 0.692]	[0.648; 0.692]	Yes/Yes	0.67	0.449		0.816
Attitude towards the destination									0.592	0.368
H1d	Credibility → Attitude towards the destination	0.175	***	8.679	[0.143; 0.209]	[0.141; 0.207]	Yes/Yes	0.608	0.106		0.041
H1g	Attitude towards the post → Attitude towards the destination	0.640	***	33.753	[0.608; 0.67]	[0.609; 0.671]	Yes/Yes	0.758	0.485		0.544
Attitude towards the post									0.493	0.456
H1c	Credibility → Attitude towards the post	0.507	***	23.286	[0.471; 0.543]	[0.469; 0.541]	Yes/Yes	0.677	0.343		0.279
H1e	Parasocial interaction → Attitude towards the post	0.253	***	11.688	[0.217; 0.288]	[0.217; 0.289]	Yes/Yes	0.593	0.150		0.069

n = 10,000 subsamples: * p < .05; ** p < .01; *** p < .001; ns: non-significant (one-tailed Student's t).

t(0.05; 4999) = 1.645 ; t(0.01; 4999) = 2.327 ; t(0.001; 4999) = 3.092.

Confidence intervals [5%–95%].

Effect sizes f²: >= 0.15 (moderate), >= 0.02 (small), < 0.02 (negligible) (Hair et al., 2022).

In the next step, we used the coefficient of determination (R²) and found that the proposed model explains 46.0% of intention to travel to the destination presented by the influencer, 41.3% of intention to recommend it, 59.2% of attitude towards the destination and 49.3% of attitude towards the post (see Table 5).

Lastly, we used the Stone-Geisser test (Geisser, 1975; Stone, 1974) to check the predictive power of the model. The results show that the Q² score is greater than 0 and therefore this criterion is met (see Table 5).

Comparison of the models

Table 6 indicates that the path coefficients are significant for each of the two models (human and virtual). In general, the human and VIs display similar patterns in terms of how their characteristics affect intention to travel, PSI and attitudes towards the post and destination. However, in several aspects, the human influencer tends to have a slightly higher impact on credibility and PSI, whilst the VI proves to have a greater impact on attitude towards destinations and intention to recommend them. These findings suggest that the conceptual model is robust and able to explain the proposed constructs.

Table 6.

Path coefficients (human vs. virtual).

Relationship		Path coefficients
Relationship		Human		Virtual
RQ2a	Credibility → Intention to travel	0.200	***	0.143	***
RQ2b	Credibility → Parasocial interaction	0.673	***	0.663	***
RQ2c	Credibility → Attitude towards the post	0.499	***	0.489	***
RQ2d	Credibility → Attitude towards the destination	0.195	***	0.157	***
RQ2e	Parasocial interaction → Attitude towards the post	0.251	***	0.253	***
RQ2f	Parasocial interaction → Intention to travel	0.182	***	0.132	***
RQ2g	Attitude towards the post → Attitude towards the destination	0.608	***	0.655	***
RQ2h	Attitude towards the destination → Intention to travel	0.416	***	0.493	***
RQ2i	Attitude towards the destination → Intention to recommend	0.613	***	0.658	***
	Coefficient of determination R2
	Intention to travel	0.466		0.450
	Intention to recommend	0.376		0.433

Significance level: *** p < 0.001; ** p < 0.01; * p < 0.05; ns: no significance.

Likewise, the coefficient of determination (R²) results show that both models, both for the human influencer and the VI, explain a significant amount of variance in intention to travel and intention to recommend. However, the human influencer model has a slightly higher R² for intention to travel, whereas the VI model has a higher R² for intention to recommend.

Discussion of results and managerial implications

The results of this research form the first contribution referring to the study of VIs in the promotion of tourism destinations. This research presents a theoretical model that analyses the effectiveness of two types of influencers – human and virtual – in decision-making regarding travel. Based on this model, we established nine research hypotheses and one research question that compared both types of influencers.

Firstly, regarding the relationships proposed in this theoretical model, we confirmed that all of them are significant, both for the human influencer and for the VI. Therefore, we consider that the proposed theoretical model is suitable for achieving the objectives set out initially. At the same time, the results of our research hypotheses are coherent with those obtained in other studies on influencers which connect the following variables: influencer credibility and intention to travel (Yılmazdoğan et al., 2021); influencer credibility and PSI (Lee and Watkins, 2016; Yılmazdoğan et al., 2021); influencer credibility and attitude towards the post (Teng et al., 2014); influencer credibility and attitude towards the destination (Shang and Luo, 2021); PSI and attitude towards the post (Gong and Li, 2017); PSI and intention to travel (Yılmazdoğan et al., 2021); attitude towards the post and attitude towards the brand/destination (Munnukka et al., 2016); attitude towards the destination and intention to travel and recommend (Alipour et al., 2020; Jalilvand et al., 2012).

Secondly, regarding the first research question proposed, we can confirm that there are differences in the variables analysed according to the type of influencer (human or virtual). In this case, the human influencer obtains higher scores in credibility, PSI, attitude towards the post, attitude towards the destination, intention to travel and recommendation of the destination. The very few studies that have compared these two types of influencers, analysing other variables, found that users respond more positively to human influencers (Arsenyan and Mirowska, 2021; Franke et al., 2023; Seymour et al., 2020).

Conclusion

In recent years, VI marketing has gained in importance; however, studies referring to its effectiveness have been scarce (Akhtar et al., 2024; Yu et al., 2024). This study has explored the effectiveness of VIs compared to human influencers in the promotion of tourism destinations and presents important findings, both theoretical and practical. Theories that have been applied in the context of influencers to enable us to understand tourist behaviour are those of PSI (e.g. Deng et al., 2022; Santateresa-Bernat et al., 2023) and source credibility (e.g. Akhtar et al., 2024; Nguyen et al., 2023).

This article is one of the first contributions to include the theory of source credibility and the theory of PSI to study the effectiveness of virtual and human influencer marketing in the tourism sector. Therefore, as the results corroborate the relationships proposed in the hypotheses put forward, the proposed theoretical model can be considered ideal for studying the choice of destination based on posts by virtual and human influencers.

In this work, we used an explanatory model that includes two theories: the theory of source credibility and the theory of PSI. The relationships proposed in the model turned out to be significant for both types of influencers, confirming the robustness of the theoretical framework used. Therefore, based on these findings, this model can be applied both to human and VIs.

The results indicate that human influencers are perceived to be more credible, generate more PSI and produce more positive attitudes towards the posts and the destinations that they are promoting, resulting in a greater intention to travel as compared to VIs. These findings suggest that human influencers have a significant advantage over VIs, although the use of VIs can also be effective when promoting a destination. In fact, it has been suggested that the use of VIs can have a number of advantages over the use of human influencers, such as greater control over the content that is published (Mrad et al., 2022), which could prevent potential brand reputation crises (Jhawar et al., 2023). Therefore, we recommend implementing a balanced marketing strategy that draws on the strengths of both types of influencers.

This work presents a series of practical implications for tourism organisations. First of all, the results of the study suggest that promoting tourist destinations on a social network like Instagram may be more effective if carried out by a human influencer. However, the actions of VIs may also be effective in destination marketing. Therefore, future research should analyse which type of influencer is the most ideal, taking into account a range of factors such as the target audience, the type of destination, the tourist activities on offer there, etc.

Secondly, analysis of source credibility has highlighted the importance of two variables: trustworthiness and attractiveness. For this reason, it would be a good idea for organisations managing tourist destination marketing to consider these aspects when collaborating with such influencers.

Thirdly, the results obtained reveal the relevance of PSI in the case of human influencers. In this sense, tourism organisations should choose human influencers for marketing that requires greater interaction with followers.

Finally, the use of VIs can provide other benefits for tourism organisations, among which we can highlight greater control over the influencer and greater personalisation of the message. So that these campaigns are effective, organisations must provide the VI with the appropriate characteristics for the target public.

However, this study does have limitations which open up future lines of research. Firstly, we were unable to propose research hypotheses in certain cases due to the scarcity of previous literature on VIs. Secondly, the theoretical model does not include variables that may be of interest in the area of influencer marketing and the promotion of tourism destinations, such as engagement, authenticity, innovation, novelty, etc. and that may explain the potential advantages of VIs over human ones that have not been identified in this study. Thirdly, the proposed theoretical model could be improved by including new relationships between the different constructs. For instance, one option would be to check whether attitude towards the post has an influence on the intention to travel and to recommend a destination (Han and Chen, 2022; Popy and Bappy, 2022). To name a fourth point, two female influencer profiles were used in the questionnaire and therefore it would be interesting to include male profiles in future studies to check if there are any differences in the variables analysed according to the influencer's sex. Finally, this study was applied specifically to the case of tourism destinations, but further research could analyse this type of campaign in other tourism organisations (airlines, hotel chains, travel agents, etc.).

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Janet Hernández-Méndez

References

Ajzen

(2001) Nature and operation of attitudes. Annual Review of Psychology 52(1): 27–58.

Akhtar

Hameed

Islam

, et al. (2024) Avatars of influence: understanding how virtual influencers trigger consumer engagement on online booking platforms. Journal of Retailing and Consumer Services 78: 103742.

Alipour

Olya

Maleki

, et al. (2020) Behavioral responses of 3S tourism visitors: evidence from a Mediterranean island destination. Tourism Management Perspectives 33: 100624.

Appel

Grewal

Hadi

, et al. (2020) The future of social media in marketing. Journal of the Academy of Marketing Science 48(1): 79–95.

Arsenyan

Mirowska

(2021) Almost human? A comparative case study on the social media presence of virtual influencers. International Journal of Human-Computer Studies 155: 102694.

Belanche

Casaló

Flavián

, et al. (2021) Understanding influencer marketing: the role of congruence between influencers, products and consumers. Journal of Business Research 132: 186–195.

Bhattacharya

(2023) Parasocial interaction in social media influencer-based marketing: an SEM approach. Journal of Internet Commerce 22(2): 272–292.

Bigne

Sanchez

(2001) Tourism image, evaluation variables and after purchase behaviour: inter-relationship. Tourism Management 22(6): 607–616.

Bower

Landreth

(2001) Is beauty best? Highly versus normally attractive models in advertising. Journal of Advertising 30(1): 1–12.

10.

Bulchand-Gidumal

(2022) Impact of artificial intelligence in travel, tourism, and hospitality. In: Handbook of e-Tourism. Cham: Springer International Publishing, 1943–1962.

11.

Chen

Shang

(2014) The effects of perceived relevance of travel blogs’ content on the behavioral intention to visit a tourist destination. Computers in Human Behavior 30: 787–799.

12.

Childers

Lemon

Hoy

(2019) # Sponsored# Ad: agency perspective on influencer marketing campaigns. Journal of Current Issues & Research in Advertising 40(3): 258–274.

13.

Chin

(2010) How to write up and report PLS analyses. In: Esposito Vinzi

Chin

Henseler

Wang

(eds) Handbook of Partial Least Squares: Concepts, Methods and Applications. Berlin, Germany: Springer-Verlag, 655–690.

14.

da Silva Oliveira

Chimenti

(2021) Humanized robots”: a proposition of categories to understand virtual influencers. Australasian Journal of Information Systems 25: 1–27.

15.

Deng

Benckendorff

Wang

(2022) From interaction to relationship: rethinking parasocial phenomena in travel live streaming. Tourism Management 93. doi:https://doi.org/10.1016/j.tourman.2022.104583

16.

Diamantopoulos

Siguaw

(2006) Formative versus reflective indicators in organizational measure development: a comparison and empirical illustration. British Journal of Management 17(4): 263–282.

17.

Drenten

Brooks

(2020) Celebrity 2.0: Lil Miquela and the rise of a virtual star system. Feminist Media Studies 20(8): 1319–1323.

18.

Elmashhara

Soares

(2022) Linking atmospherics to shopping outcomes: the role of the desire to stay. Journal of Retailing and Consumer Services 64: 102744.

19.

Faul

Erdfelder

Buchner

, et al. (2009) Statistical power analyses using G*Power 3.1: tests for correlation and regression analyses. Behavior Research Methods 41: 1149–1160.

20.

Franke

Groeppel-Klein

Müller

(2023) Consumers’ responses to virtual influencers as advertising endorsers: novel and effective or uncanny and deceiving? Journal of Advertising 52(4): 523–539. DOI: https://doi.org/10.1080/00913367.2022.2154721.

21.

Geisser

(1975) The predictive sample reuse method with applications. Journal of the American Statistical Association 70(350): 320–328.

22.

Gold

Malhotra

Segars

(2001) Knowledge management: an organizational capabilities perspective. Journal of Management Information Systems 18(1): 185–214.

23.

Gong

(2017) Engaging fans on microblog: the synthetic influence of parasocial interaction and source characteristics on celebrity endorsement. Psychology & Marketing 34(7): 720–732.

24.

Gretzel

(2018) Influencer marketing in travel and tourism. In: Sigala

Gretzel

(eds) Advances in Social Media for Travel, Tourism and Hospitality: New Perspectives, Practice and Cases. New York: Routledge, 147–156.

25.

Hair

Hult

GTM

Ringle

, et al (2022) A Primer on Partial Least Squares Structural Equation Modeling (PLS-SEM). 3rd ed. Thousand Oaks: Sage.

26.

Hair

Ringle

Sarstedt

(2011) PLS-SEM: indeed a silver bullet. Journal of Marketing Theory and Practice 19(2): 137–149.

27.

Hair

Risher

Sarstedt

, et al (2019) When to use and how to report the results of PLS-SEM. European Business Review 31(1): 2–24.

28.

Han

Chen

(2022) Millennial social media users’ intention to travel: the moderating role of social media influencer following behavior. International Hospitality Review 36(2): 340–357.

29.

Henseler

Hubona

Ray

(2016) Using PLS path modeling in new technology research: updated guidelines. Industrial Management & Data Systems 116(1): 2–20.

30.

Henseler

Ringle

Sinkovics

(2009) The use of partial least squares path modeling in international marketing. Advances in International Marketing 20: 277–319.

31.

Horton

Richard Wohl

(1956) Mass communication and para-social interaction: observations on intimacy at a distance. Psychiatry 19(3): 215–229.

32.

Jalilvand

Samiei

Dini

, et al. (2012) Examining the structural relationships of electronic word of mouth, destination image, tourist attitude toward destination and travel intention: an integrated approach. Journal of Destination Marketing & Management 1(1-2): 134–143.

33.

Jhawar

Kumar

Varshney

(2023) The emergence of virtual influencers: a shift in the influencer marketing paradigm. Young Consumers 24(4): 468–484.

34.

Jin

Ryu

(2020) I'll buy what she's# wearing”: the roles of envy toward and parasocial interaction with influencers in Instagram celebrity-based brand endorsement and social commerce. Journal of Retailing and Consumer Services 55: 102121.

35.

Jin

Ryu

Muqaddam

(2021) I trust what she's# endorsing on Instagram: moderating effects of parasocial interaction and social presence in fashion influencer marketing. Journal of Fashion Marketing and Management: An International Journal 25(4): 665–681.

36.

Kim

(2022) Keeping up with influencers: exploring the impact of social presence and parasocial interactions on Instagram. International Journal of Advertising 41(3): 414–434.

37.

Lee

Watkins

(2016) Youtube vloggers’ influence on consumer luxury brand perceptions and intentions. Journal of Business Research 69(12): 5753–5760.

38.

Lee

Kim

(2020) Influencer marketing on Instagram: how sponsorship disclosure, influencer credibility, and brand credibility impact the effectiveness of Instagram promotional post. Journal of Global Fashion Marketing 11(3): 232–249.

39.

Leung

Palmatier

(2022) Online influencer marketing. Journal of the Academy of Marketing Science 50(2): 226–251.

40.

Lin

Crowe

Pierre

, et al. (2021) Effects of parasocial interaction with an instafamous influencer on brand attitudes and purchase intentions. The Journal of Social Media in Society 10(1): 55–78.

41.

Lou

Kiew

STJ

Chen

, et al. (2023) Authentically fake? How consumers respond to the influence of virtual influencers. Journal of Advertising 52(4): 540–557.

42.

Miquela [@lilmiquela]. Instagram. https://www.instagram.com/lilmiquela/.

43.

Mitchell

Olson

(1981) Are product attribute beliefs the only mediator of advertising effects on brand attitude? Journal of Marketing Research 18(3): 318–332.

44.

Mrad

Ramadan

Nasr

(2022) Computer-generated influencers: the rise of digital personalities. Marketing Intelligence & Planning 40(5): 589–603.

45.

Munnukka

Uusitalo

Toivonen

(2016) Credibility of a peer endorser and advertising effectiveness. Journal of Consumer Marketing 33(3): 182–192.

46.

Nguyen

PMB

Pham

Truong

GNT

(2023) The influence of source credibility and inspiration on tourists’ travel planning through travel vlogs. Journal of Travel Research: 00472875231206538.

47.

Ohanian

(1990) Construction and validation of a scale to measure celebrity endorsers’ perceived expertise, trustworthiness, and attractiveness. Journal of Advertising 19(3): 39–52.

48.

Penttinen

Ciuchita

Čaić

(2022) Youtube it before you buy it: the role of parasocial interaction in consumer-to-consumer video reviews. Journal of Interactive Marketing 57(4): 561–582.

49.

Podsakoff

Organ

(1986) Self-reports in organizational research: problems and prospects. Journal of Management 12(4): 531–544.

50.

Popy

Bappy

(2022) Attitude toward social media reviews and restaurant visit intention: a Bangladeshi perspective. South Asian Journal of Business Studies 11(1): 20–44.

51.

Ringle

Wende

Becker

J-M

(2022) SmartPLS 4. Oststeinbek: SmartPLS GmbH, http://www.smartpls.com .

52.

Sands

Campbell

Plangger

, et al. (2022b) Unreal influence: leveraging AI in influencer marketing. European Journal of Marketing 56(6): 1721–1747.

53.

Sands

Ferraro

Demsar

, et al. (2022a) False idols: unpacking the opportunities and challenges of falsity in the context of virtual influencers. Business Horizons 65(6): 777–788. https://doi.org/10.1016/j.bushor.2022.08.002 .

54.

Santateresa-Bernat

Sánchez-García

Curras-Perez

(2023) I like you, or I like what you say? Effect of influencer on tourists’ behaviours. Current Issues in Tourism 26(19): 3160–3174.

55.

Sarstedt

Hair

Ringle

, et al. (2016) Estimation issues with PLS and CBSEM: Where the bias lies!. Journal of business research 69(10): 3998–4010.

56.

Seymour

Yuan

Dennis

, et al. (2020) Facing the artificial: understanding affinity, trustworthiness, and preference for more realistic digital humans. In: Proceedings of the 53rd Hawaii International Conference on System Sciences. https://doi.org/10.24251/HICSS.2020.574 .

57.

Shang

Luo

(2021) Modeling of the impact of the credibility of the destination endorser on the place attachment of potential tourists. Frontiers in Psychology 12: 759207.

58.

Sokolova

Kefi

(2020) Instagram and YouTube bloggers promote it, why should I buy? How credibility and parasocial interaction influence purchase intentions. Journal of Retailing and Consumer Services 53: 101742.

59.

Stone

(1974) Cross-validation and multinomial prediction. Biometrika 61(3): 509–515.

60.

Teng

Wei Khong

Wei Goh

, et al. (2014) Examining the antecedents of persuasive eWOM messages in social media. Online Information Review 38(6): 746–768.

61.

Thomas

Fowler

(2021) Close encounters of the AI kind: use of AI influencers as brand endorsers. Journal of Advertising 50(1): 11–25.

62.

Time (2018) The 25 Most Influential People on the Internet https://time.com/5324130/most-influential-internet/ (acceso 19/01/2023).

63.

Tussyadiah

(2020) A review of research into automation in tourism: launching the annals of tourism research curated collection on artificial intelligence and robotics in tourism. Annals of Tourism Research 81: 102883.

64.

Weismueller

Harrigan

Wang

, et al. (2020) Influencer endorsements: how advertising disclosure and source credibility affect consumer purchase intention on social media. Australasian Marketing Journal 28(4): 160–170.

65.

Yılmaz

Sezerel

Uzuner

(2020) Sharing experiences and interpretation of experiences: a phenomenological research on Instagram influencers. Current Issues in Tourism 23(24): 3034–3041.

66.

Yılmazdoğan

Doğan

RŞ

Altıntaş

(2021) The impact of the source credibility of Instagram influencers on travel intention: the mediating role of parasocial interaction. Journal of Vacation Marketing 27(3): 299–313.

67.

Dickinger

KKF

, et al. (2024) Artificial intelligence-generated virtual influencer: examining the effects of emotional display on user engagement. Journal of Retailing and Consumer Services 76: 103560.

68.

Yuan

Lou

(2020) How social media infuencers foster relationships with followers: the roles of source credibility and fairness in parasocial relationship and product interest. Journal of Interactive Advertising 20(2): 133–147.

69.

Zhang

Gursoy

(2020) The effect of celebrity endorsement on destination brand love: a comparison of previous visitors and potential tourists. Journal of Destination Marketing & Management 17: 100454.