Creator Trustworthiness Signals: What to Track Before You Partner by Manalt Agency

Proxies for Creator Trust

Which observable social-platform signals (e.g., comment quality, sentiment, share/save rate, engagement consistency, sponsored-post performance gap) are the strongest proxies for perceived creator trustworthiness?

The sponsored-organic performance gap serves as the most consistent quantifable trust proxy across contexts, while followership status and high-arousal language are strongest for micro-infuencers, expertise demonstration and argument quality are strongest for macro-infuencers, and perceived interaction dominates in live-streaming contexts—with traditional engagement metrics functioning primarily as downstream consequences of trustworthiness rather than direct proxies.

Abstract

The strongest proxies for perceived creator trustworthiness vary systematically by infuencer audience size and platform context. For micro-infuencers, followership status emerges as the most powerful signal (Δ = 1.86–1.91, p < 0.01), followed by high-arousal language (IRR = 1.036, p < .001) and clear sponsorship disclosures (r = 0.42 with perceived authenticity). For macro-infuencers, expertise demonstration is the strongest predictor (²= .36, p < .001), with argument quality, transparent disclosure practices, and brand-infuencer congruence serving as important secondary signals. The sponsored-organic performance gap functions as a universal, quantifable trust proxy across contexts, with sponsored videos costing infuencers 0.17–0.19% of their subscriber base compared to equivalent organic content, though this efect is amplifed for larger audiences and mitigated by content ft with the creator's usual material.

Platform architecture conditions signal efectiveness: search-driven platforms favor mixed sentiment as a credibility signal, while scroll-driven platforms favor positive sentiment as a likability signal. In live-streaming contexts, perceived interaction moderated by homophily (²= 0.176, p < .01) supersedes expertise signals (²= −0.137, p < .05) as the primary trust proxy. For expertise-oriented domains, account longevity (· ² = 0.14), content consistency, and network status serve as strong trust indicators. Audience usage intensity also moderates signal efectiveness, with heavy platform users responding positively to disclosure signals that light users do not register. Traditional engagement metrics (likes, comments, shares) function primarily as downstream consequences of trustworthiness rather than direct proxies, with their interpretation dependent on audience size and content context.

Screening

We screened in sources based on their abstracts that met these criteria:

• Digital Platform Context: Does this study examine social media platforms or digital content platforms where creator-audience relationships develop?

• Trustworthiness Outcome: Does this study measure or assess perceived trustworthiness, credibility, or reliability of content creators, infuencers, or social media users?

• Platform Signals and Relationships: Does this study analyze observable behavioral or engagement metrics on social platforms AND establish or test relationships between these platform signals and trustworthiness perceptions?

• Human Participants: Does this study involve human participants as evaluators or audiences making trustworthiness assessments?

• Study Design: Is this an empirical study (experimental, quasi-experimental, cross-sectional, longitudinal, or observational design) or a systematic review/meta-analysis with empirical data?

• Creator Focus: Does this study focus on creator trustworthiness rather than solely on information content trustworthiness or accuracy?

• Behavioral Signals: Does this study include behavioral or engagement signals rather than focusing exclusively on static demographic or profle characteristics?

• Generalizability: Does this study provide broader generalizability beyond single-creator case studies and include empirical data rather than being purely theoretical?

We considered all screening questions together and made a holistic judgement about whether to screen in each paper.

TLDR: The Behavioral Proxy Approach to Creator Trustworthiness

1. The Paradigm Shift: From Metrics to Behavioral Proxies

In the mature creator economy, traditional Key Performance Indicators (KPIs) such as likes, shares, and raw engagement rates have become noisy variables. From a behavioral science perspective, these metrics represent downstream consequences. They are the secondary efects of an audience’s pre-existing trust, not the upstream predictors of it.

The data supports a shift away from monitoring lagging indicators and toward identifying behavioral proxies. These are observable signals that reveal the underlying structural integrity of a creator’s infuence.

Relying on traditional engagement metrics creates “reputation-burning” partnerships. In this dynamic, a creator spends down their social capital to fulfll sponsorship obligations, which drives long-term decay of both creator equity and brand equity. A proxy-based approach mitigates this risk by prioritizing source credibility markers that persist beyond a single campaign.

Metric Category Examples Strategic Classifcation

Traditional Engagement

Behavioral Trust Proxies

Likes, Comments, Shares, Saves

Performance gaps, linguistic arousal, argument quality

Lagging indicators. Downstream consequences of trust.

Leading behavioral indicators. Predictive signals of underlying integrity.

High engagement on a sponsored post can mask an authenticity crisis. When strategists optimize for raw reach and surface-level metrics, they risk entering an adverse selection loop where they partner with creators who are efectively cashing out their reputation. By prioritizing leading proxies, high-value brands can select creators whose credibility ensures the message is internalized, not just viewed.

2. The Universal Signal: Analyzing the Sponsored-Organic Performance Gap

The most consistent quantifable indicator of creator trustworthiness across digital ecosystems is the sponsored-organic performance gap. This delta captures the “reputation-burning efect” that emerges when commercial content reduces perceived authenticity.

Evidence across YouTube and Bilibili shows that sponsored videos typically cost infuencers between 0.17% and 0.19% of their subscriber base compared to equivalent organic content.

This gap provides a direct diagnostic window into the Persuasion Knowledge Model. A high delta signals that the audience has interpreted the content as a commercial disruption rather than a value contribution, triggering defensive processing.

The performance gap is not inevitable. Strategists can reduce the reputation-burning efect through two control levers:

1. Brand-infuencer congruenceAlign the sponsorship with the creator’s usual thematic material.

2. Brand selectionPromoting less well-known brands can reduce the performance gap, since audiences may interpret the partnership as authentic discovery rather than pure commercialization.

In addition, face presence, particularly in video formats, operates as a meaningful moderator that softens negative instantaneous audience response such as live comments and sentiment.

3. Segmented Vetting: Micro-Influencer vs. Macro-Influencer Trust Signals

Vetting protocols must be calibrated to creator audience size. The psychological flters applied by audiences change as a creator scales, refected in the “authenticity premium” being materially stronger for micro-infuencers (²= 0.63) than for mega-infuencers (²= 0.41).

Vetting Category

Primary Proxy

Linguistic Signal

Engagement Driver

Micro-Infuencers (<100k Followers)

Followership status. Strongest signal of trust (Δ = 1.86 to 1.91).

High-arousal language. Enthusiasm functions as an authenticity marker (IRR = 1.036).

Parasocial interaction. Relatability and responsiveness drive trust.

Misapplying these signals is a frequent strategic failure.

Macro-Infuencers (>100k Followers)

Expertise demonstration. Core predictor of credibility (²= .36).

Formal language markers. Articles, prepositions, and long-form structure signal authority.

Informative goal framing. Logical, objective framing increases engagement by 1.8%.

• High-arousal enthusiastic language increases trust for micro-infuencers but reduces impact for macro-infuencers (IRR = .944).

• For large creators, intense enthusiasm is often decoded as commercially motivated exaggeration. For macro-infuencers, the data supports a stronger focus on argument quality and counterbalanced valence, meaning the creator explicitly addresses both strengths and limitations. This stabilizes credibility and reduces perceived manipulation.

4. Environmental Conditioning: Platform Architecture and Content Format

Trustworthiness is not a static trait. It is conditioned by platform afordances, meaning the mechanisms through which each platform delivers and ranks content. The audience mindset difers between search-oriented and scroll-oriented environments, which shifts the credibility flter applied to creator signals.

• Scroll-driven platforms (TikTok, Instagram)Users operate in an experiential mindset where positive sentiment drives follow behavior. Sociable language can mitigate the platform preference for positivity, enabling more nuance without eroding likability.

• Search-driven platforms (Yelp, Goodreads)Users operate in a goal-directed mindset. Mixed sentiment becomes the primary credibility marker, while overly positive reviews can be discounted as inauthentic.

• Live-streaming environments (Taobao Live, Twitch)Relational signals dominate competence signals. Perceived interaction and homophily, meaning perceived similarity, become the primary trust proxies (²= 0.176).

Live-streaming contains a major execution trap. Homophily negatively moderates the expertise-trust relationship (²= −0.137). This means that in high-relatability contexts, an overly formal expert posture can reduce trust.

Strategists should also evaluate companion presence, such as others appearing in-frame or visible social context cues. These increase perceived humanness and signifcantly boost trustworthiness (IRR = 1.699).

5. The Disclosure Framework: Managing the Transparency Paradox

The transparency paradox describes the tension between two proven mechanisms:

• Persuasion Knowledge Model explains immediate negative reactions to explicit disclosures such as “#ad”

• Source credibility explains the long-term value of transparency in sustaining trust

Disclosure therefore has a dual time-horizon efect:

• Immediate efectsDisclosed ads may experience lower instantaneous engagement.

• Persistent efectsTransparent disclosure produces positive persistent efects on future engagement, while undisclosed advertising creates long-term credibility decay.

Strategic Implication

Disclosure management must be segmented by audience usage intensity.

• Heavy platform users respond positively to clear “#ad” disclosures (Δ = 0.443) because transparency increases perceived integrity.

• Light users often do not register disclosure signals and show minimal response.

In addition, platform-initiated branded content tools tend to generate stronger negative efects (−0.254) than creator-authored in-text disclosures. The recommended approach is to use both. This maintains compliance while allowing the creator’s own wording to soften the standardized platform tag.

6. Implementation: The Behavioral Vetting Protocol

To move from measuring reach to managing reputation, a multi-layered behavioral vetting protocol should be applied.

1. Longevity and content consistency auditEvaluate account registration age (· ² = 0.14 for credibility) and thematic consistency. Consistency in topic signature over time is a proven credibility predictor.

2. Linguistic arousal calibrationEnsure the creator’s tone matches their audience scale. High arousal is efective for micro-infuencers; informational framing and controlled language is more efective for macro-infuencers.

3. Persistence and sincerity checkReview historical audience feedback dynamics. Creators who disable comments are perceived as less receptive to consumer voice and therefore less sincere.

4. Network oversight and status verifcationFor expertise-led domains, confrm external validation signals such as contributor publication experience and ownership status in specialized networks.

Satisfcing theory suggests audiences do not require a perfect score across all trust signals. A brand does not need to optimize every proxy simultaneously.

Instead, identify one “hero signal” that anchors credibility, such as:

• strong longevity track record

• consistently low sponsored-organic performance gap

• exceptional argument quality

A deep audit for one high-impact anchor is strategically superior to a broad but shallow checklist approach.

Our take

The transition from engagement-based metrics to behavioral proxies is a shift from reactive monitoring to proactive reputation management. By prioritizing the true signals of trust, including the sponsored-organic performance gap, segment-specifc linguistic cues, and platform-conditioned sentiment dynamics, brands can build creator partnerships that persist beyond the current algorithm cycle.

Characteristics of Included Studies

This systematic review synthesizes fndings from 40 studies examining observable social-platform signals as proxies for creator trustworthiness. The included studies span multiple platforms, methodological approaches, and geographic contexts. See appendix

The included studies demonstrate methodological diversity, with 15 purely observational designs, 9 experimental studies, and 10 mixed-methods approaches. Instagram was the most frequently studied platform (14 studies), followed by YouTube (8 studies) and TikTok (4 studies). Several studies examined enterprise or specialized platforms such as social trading platforms and investment advice platforms. Geographic coverage includes North America, Europe, and Asia, though many studies did not specify location. Full texts were available for 16 of the 40 studies, with the remainder analyzed based on abstracts only.

Engagement Metrics as Trust Proxies

Multiple studies examined traditional engagement metrics (likes, comments, shares) as signals of trustworthiness, though fndings reveal these metrics serve more as consequences than proxies of trust. Zhang et al. (2023) found that engagement metrics including likes and comments difered signifcantly between sponsored and organic content, with sponsored videos generating weaker audience responses. Similarly, Cheng et al. (2024) observed that audience response gaps in likes, comments, and comment texts were larger among infuencers with larger audiences, suggesting engagement patterns signal authenticity concerns rather than directly predicting trustworthiness.

Signal Category

Engagement metrics

Specifc Metrics

Likes, comments, shares

Likes, comments

Engagement metrics Follower count

Engagement metrics

Content consumption, contribution

Direction of Efect Study Context

Indirect proxy via response gaps YouTube beauty/style

Lower for sponsored vs. organic content YouTube beauty/style

Complex relationship with trustworthiness Instagram, TikTok

Mediated by parasocial relationships Instagram

Cascio Rizzo et al. (2023) demonstrated that engagement serves as a downstream indicator of trustworthiness rather than an antecedent, with high-arousal language increasing engagement for micro-infuencers (IRR = 1.036, p < .001) but decreasing engagement for macro-infuencers (IRR = .944, p < .001). This pattern suggests that engagement metrics must be interpreted in light of audience size and content characteristics.

Content Characteristics and Language Signals

Content-based signals emerged as among the strongest predictors of perceived trustworthiness across studies. Argument quality was identifed as a particularly important signal, especially for larger infuencers. Pozharliev et al. (2022) found that meso-infuencers (10,000 to 1 million followers) were perceived as credible sources only when their posts provided strong argument quality, with trustworthiness measured using Ohanian's scale achieving ±= 0.95.

Content Signal Efect on Trustworthiness Efect Size/Signifcance Infuencer Type

Argument quality (strong) Positive for meso-infuencers Signifcant (EEG + self-report) Meso (10K-1M)

High-arousal language

Informative goal framing

Counterbalanced valence

Post objectivity

Positive for micro, negative for macro IRR = 1.036 (micro), 0.944 (macro), p < .001

Mitigates negative efects for macro

Positive for macro-infuencers

Positive association with authenticity

34% stronger signal → 1.8% engagement increase, p = .033

Signifcant

Signifcant via SEM

Micro vs. macro

Macro

Non-celebrity SMIs

Han et al. (2018) identifed specifc linguistic features predictive of author credibility on Twitter: positive associations with account registration age, tweets similarity, sentiments in tweets, formal language markers (articles, prepositions, six-letter words), with efect sizes measured through ² values (e.g., ² = 0.14 for account registration age). Negative associations were found for swear words, frst-person singular pronouns, questions, and negations.

Aggregate sentiment emerged as a cross-platform indicator of credibility. Shalev et al. (2025) found that aggregate communicator sentiment across multiple posts diferentially afected following decisions depending on platform type: positive sentiment drove following on scroll-driven platforms (Twitter/X, Instagram), while mixed sentiment signaled credibility on search-driven platforms (Yelp, Goodreads).

Creator-Audience Interaction Signals

Responsiveness and interaction quality emerged as signifcant trust predictors, particularly in social commerce contexts. Senali et al. (2024) found that responsiveness positively infuenced trust in sellers on Instagram, while review quantity and review quality both served as positive predictors of trust.

Interaction Signal Efect Direction

Responsiveness Positive

Perceived interaction Positive, moderated by homophily

Interactivity (platform) Positive for trust in streamers

Network status Positive

Context

Measurement

Instagram s-commerce PLS analysis

Taobao livestream ²= 0.176, p < .01

Live-streaming e-commerce SEM

Social trading 38-week longitudinal

Comment disabling Negative (reduced sincerity) Twitter Experimental

Daniels et al. (2024) found that infuencers who disabled social media comments were perceived as less receptive to consumer voice and thus less sincere, leading to more negative impressions and reduced persuasiveness. This efect was mitigated when consumers believed self-protection was a reasonable priority.

In live-streaming contexts, Cao et al. (2025) found that perceived interaction positively predicted trust, with this relationship strengthened by homophily (²= 0.176, p < .01). Notably, homophily negatively moderated the expertise-trust relationship (²= −0.137, p < .05), suggesting that when audiences perceive similarity with streamers, interaction signals become more important than expertise signals.

Visual and Authenticity Markers

Visual signals related to creator presence demonstrated signifcant efects on trustworthiness perceptions. Ma et al. (2023) found that infuencers showing their faces during advertisements moderated the negative efects on instantaneous audience response, with signifcant changes in live comments and sentiment (p < 0.001).

Cascio Rizzo et al. (2023a) examined virtual infuencers and found that companion presence in photos signifcantly boosted perceived trustworthiness (IRR = 1.699, SE = .068, t = 13.24, p < .001). This efect operated through increased perceptions of humanness, which enhanced trust.

Brand-infuencer congruence served as a consistent authenticity signal. Cheng et al. (2022, 2024) found that high ft between sponsored content and an infuencer's "usual" content mitigated the reputation-burning efect. Similarly, promoting less well-known brands reduced negative efects on trustworthiness.

Expertise and Source Credibility Signals

Expertise demonstration emerged as a strong trust predictor across multiple contexts. Van Canh Vu et al. (2024) found expertise to be the strongest predictor of trustworthiness among source credibility dimensions (²= .36, p < .001), followed by attitude homophily (²= .20, p < .001), social attractiveness (²= .19, p < .001), and physical attractiveness (²= .09, p < .05).

Credibility Dimension ²Coefcient Signifcance Study

Expertise

Attitude homophily

²= .36

²= .20

p < .001

Van Canh Vu et al., 2024

Social attractiveness

= .19

< .001

Physical attractiveness ²= .09 p < .05

Content expertise (virtual infuencers)

Personalization

Positive efect

Positive for product trust

Signifcant

Van Canh Vu et al., 2024

Jihye Kim et al., 2024

Tedjakusuma et al., 2025

In investment contexts, Wang et al. (2014) found that a subset of "top authors" on SeekingAlpha contributed content showing signifcantly higher correlation with future stock performance, and these authors could be identifed through user interactions with their articles without requiring historical market data. This suggests that audience engagement patterns can serve as proxies for expertise-based trustworthiness.

Elliott et al. (2018) found that network oversight, contributor publication experience, and contributor ownership status all served as positive credibility cues for investors. Importantly, the presence of one positive credibility cue was sufcient to increase investor perceptions, with additional cues having little incremental infuence—a pattern consistent with satisfcing theory.

Behavioral Consistency Signals

Account longevity and behavioral consistency emerged as trust signals in several studies. Han et al. (2018) found account registration age to be among the strongest predictors of perceived credibility on Twitter, with ² = 0.14. Tweet similarity (consistency in content themes) also showed positive associations with credibility perceptions.

Wilczyński et al. (2025) found that frequent information-seeking from infuencers (OR = 3.54, 95% CI 2.45–5.22) was the strongest predictor of trust among physiotherapy students, followed by perceiving infuencers as more informative than academic staf (OR = 2.00, 95% CI 1.46–2.76) and intensive Instagram use (OR = 1.41, 95% CI 1.06–1.87). Age, study year, and prior critical-appraisal training were not signifcant predictors.

Followership status demonstrated particularly strong efects. Saternus et al. (2022) found that being a follower of an infuencer strongly improved trustworthiness perceptions (Δ = 1.909 for heavy users, Δ = 1.861 for light users, p < 0.01), making it the strongest observed proxy in their study compared to disclosure type efects.

Synthesis

The heterogeneity in fndings across studies can be explained through several key distinctions related to infuencer characteristics, content context, and audience factors.

Influencer Size as a Critical Moderator

The most consistent pattern across studies is that the relationship between social-platform signals and trustworthiness depends fundamentally on infuencer audience size. Studies consistently show that micro-infuencers and macro-infuencers operate under diferent signal-trust dynamics.

For micro-infuencers, high-arousal language increases perceived trustworthiness and engagement (IRR = 1.036, p < .001), while the same language decreases trustworthiness for macro-infuencers (IRR = .944, p < .001). This pattern can be explained mechanistically: micro-infuencers maintain closer parasocial relationships with smaller audiences who interpret enthusiasm as authentic excitement, whereas larger audiences of macro-infuencers interpret the same signals as commercially motivated exaggeration.

Similarly, the reputation-burning efect from sponsored content is signifcantly stronger among infuencers with larger audiences. Zhang et al. (2023) found that smaller audiences form stronger ties and are more receptive to

sponsored content, while Steils et al. (2022) found that disclosure messages improve engagement for macro-infuencers only when published by the infuencer rather than the platform.

Dzreke et al. (2025) quantifed this diference, fnding that micro-infuencers showed a stronger authenticity premium (² authenticity = 0.63) compared to mega-infuencers (² authenticity = 0.41). This suggests that for micro-infuencers, authenticity signals (relatability, genuine enthusiasm) are the primary trust proxies, while for macro-infuencers, expertise signals and transparent disclosure become more important.

Platform Type Conditions Signal Effectiveness

The efectiveness of specifc signals varies by platform architecture and user goals. Shalev et al. (2025) demonstrated that search-driven platforms (Yelp, Goodreads) favor credibility-signaling behaviors, where mixed sentiment serves as a trust proxy, while scroll-driven platforms (Twitter/X, Instagram) favor likability-signaling behaviors, where positive sentiment drives following. This distinction refects diferent user orientations: goal-directed information seeking versus experiential content consumption.

Alternative signals can substitute for primary trust cues depending on platform context. On scroll-driven platforms, sociable language mitigates the preference for positive sentiment. On search-driven platforms, formal credibility badges (e.g., Yelp's "Elite" status) reduce reliance on mixed sentiment as a credibility signal.

In live-streaming contexts, real-time interaction signals become paramount. Cao et al. (2025) found that perceived interaction's efect on trust was strengthened by homophily (²= 0.176, p < .01), while expertise's efect was weakened (²= −0.137, p < .05). This suggests that in synchronous, interactive contexts, relational signals supersede competence signals as trust proxies.

Content Context and Commercial Intent

The commercial context of content fundamentally shapes signal-trust relationships. Multiple studies demonstrate that sponsored content triggers persuasion knowledge activation, altering how audiences interpret signals.

Waltenrath et al. (2024) reconciled seemingly contradictory fndings about disclosure efects by distinguishing immediate from persistent efects. Disclosed advertising gathers less immediate engagement than undisclosed advertising, consistent with persuasion knowledge activation. However, transparent disclosure produces positive persistent efects on future engagement, while undisclosed advertising produces negative persistent efects. This suggests that source credibility explains long-term efects, while the Persuasion Knowledge Model explains immediate responses.

Brand characteristics also moderate signal efectiveness. Less well-known brands mitigate the reputation-burning efect, and Walsh et al. (2024) found that for popular creators, sponsorship can enhance consumer engagement when the endorsed brand is perceived as small. This suggests that audiences interpret endorsements of smaller brands as more authentic choices rather than purely commercial arrangements.

Audience Characteristics and Usage Intensity

Audience characteristics condition how signals are interpreted. Saternus et al. (2022) found that heavy Instagram users (ten or more hours per week) responded positively to "#ad" disclosures (Δ = 0.443, p < 0.05), while light users showed no signifcant response. This suggests that experienced users value transparency signals that less experienced users may not notice or prioritize.

Trust disposition negatively moderates signal-trust relationships. Senali et al. (2024) found that trust disposition negatively moderated the impact of review quality on trust in sellers and responsiveness on trust in products. High-trust individuals may rely less on external signals because they approach interactions with favorable baseline assumptions.

Age demonstrates an inverse correlation with susceptibility to infuencer persuasion, though Wilczyński et al. (2025) found that age was not a signifcant predictor of trust in physiotherapy infuencers among students. This apparent contradiction may refect that age efects operate through diferent mechanisms in professional versus consumer contexts.

Strongest Signal Proxies by Context

Based on the synthesis of efect sizes and consistency across studies, the following hierarchy of signal strength emerges:

For micro-infuencers: Followership status (Δ = 1.86–1.91, p < 0.01), high-arousal language (IRR = 1.036), authenticity cues (r = 0.42 for clear disclosures), and relational signals (parasocial interaction, responsiveness).

For macro-infuencers: Expertise demonstration (²= .36, p < .001), transparent disclosure (positive persistent efects), informative framing (1.8% engagement increase), argument quality (signifcant EEG and self-report efects), and brand-infuencer congruence.

For live-streaming contexts: Perceived interaction (²= 0.176 with homophily), interactivity features, face presence (p < 0.001 for sentiment moderation).

For investment/expertise domains: Account longevity (² = 0.14), content consistency, network status, publication experience, and user interaction patterns with content.

The sponsored-organic performance gap serves as a universal, quantifable signal across contexts, with consistent efect sizes of 0.17–0.19% reputation cost per sponsored video on YouTube, though this efect is moderated by audience size, content ft, and brand characteristics.

References

A. P. Tedjakusuma, Andri Dayarana K. Silalahi, Ixora Javanisa Eunike, Do Thi Thanh Phuong, and Dalianus Riantama. “The Trust-Driven Path to Consumer Engagement Behaviors: Exploring the Role of Streamer and Platform Characteristics in Live-Streaming E-Commerce.” Digital Business, 2025.

A. Rynarzewska, Lubna Nafees, and A. Nikolov. “Typology and Impact of Follower Responses to Social Media Infuencer “Promotional Crisis”.” Psychology &Amp; Marketing, 2024.

Adrian Waltenrath. “Consumers’ Ambiguous Perceptions of Advertising Disclosures in Infuencer Marketing: Disentangling the Efects on Current and Future Social Media Engagement.” Electronic Markets, 2024.

Ágnes Buvár, Eszter Balogh, Attila Marcelián Dobai, and Judit Tessényi. “Do You Care to Share? The Negative Efect of Sponsored Content on the Behavioral Engagement with a Social Media Post Containing a COVID-19 Message.” Journal of Interactive Advertising, 2024.

Anmol Bansal. “Bridging the Perception Gap: An Investigation into Consumer Trust and Brand Value in India's Infuencer Ecosystem.” International Journal of Business and Economics Research, 2025.

Bartosz Wilczyński, Hubert Antczak, Karol de Tillier, Marcin Taraszkiewicz, Krzysztof Sobczak³, and K. Zorena. “Physiotherapy Students’ Trust in Social-Media Physiotherapy Infuencers: Implications for Digital-Literacy Training in Medical Education.” BMC Medical Education, 2025.

D. C. Hugh, Rebecca Dolan, P. Harrigan, and H. Gray. “Infuencer Marketing Efectiveness: The Mechanisms That Matter.” European Journal of Marketing, 2022.

Darlene Walsh, Argiro Kliamenakis, Michel Laroche, and Sarah Jabado. “Authenticity in TikTok: How Content Creator Popularity and Brand Size Infuence Consumer Engagement with Sponsored Usergenerated Content.” Psychology &Amp; Marketing, 2024.

Edith Shalev, Coby Morvinski, Meyrav Shoham, and Ellie J Kyung. “Search or Scroll: How Credibility Versus Likability Premiums Shape Consumer Following Decisions.” Journal of Consumer Research, 2025.

Estefania Ballester, Carla Ruiz, Natalia Rubio, and Cleopatra Veloutsou. “We Match! Building Online Brand Engagement Behaviours Through Emotional and Rational Processes.” Journal of Retailing and Consumer Services, 2025.

G. Wang, Tianyi Wang, Bolun Wang, Divya Sambasivan, Zengbin Zhang, Haitao Zheng, and Ben Y. Zhao. “Crowds on Wall Street: Extracting Value from Social Investing Platforms.” arXiv.org, 2014.

Giovanni Luca Cascio Rizzo, Francisco Javier Villarroel Ordenes, Rumen Pozharliev, Matteo De Angelis, and Michele Costabile. “How High-Arousal Language Shapes Micro- Versus Macro-Infuencers’ Impact.” Journal of Marketing, 2023.

Giovanni Luca Cascio Rizzo, Jonah A. Berger, and F. Villarroel. “What Drives Virtual Infuencer's Impact?” Social Science Research Network, 2023.

Giuseppe Soda, Aks Zaheer, Michael Park, Bill McEvily, and Mani Subramani. “Prismatic Trust: How Structural and Behavioral Signals in Networks Explain Trust Accumulation.” Management Sciences, 2024.

Jihye Kim, Minseong Kim, and Sae-Mi Lee. “Unlocking Trust Dynamics: An Exploration of Playfulness, Expertise, and Consumer Behavior in Virtual Infuencer Marketing.” International Journal of Human Computer Interactions, 2024.

Komal Shamim, and Tahir Islam. “Digital Infuencer Marketing: How Message Credibility and Media Credibility Afect Trust and Impulsive Buying.” Journal of Global Scholars of Marketing Science, 2022.

Kyungsik Han. “How Do You Perceive This Author? Understanding and Modeling Authors’ Communication Quality in Social Media.” PLoS ONE, 2018.

M. Cheng, and Shunyuan Zhang. “Reputation Burning: Analyzing the Impact of Brand Sponsorship on Social Infuencers.” Social Science Research Network, 2022.

———. “Reputation Burning: Analyzing the Impact of Brand Sponsorship on Social Infuencers.” Management Sciences, 2024.

M. Ghosh, and A. Islam. “The Efect of Social Media Infuencers’ Content Characteristics on Millennial Consumers’ Engagement on Facebook Brand Pages.” Journal of Contemporary Marketing Science, 2025.

Madugoda Gunaratnege Senali, Mohammad Iranmanesh, Morteza Ghobakhloo, B. Foroughi, Shahla Asadi, and Abderahman Rejeb. “Determinants of Trust and Purchase Intention in Social Commerce: Perceived Price Fairness and Trust Disposition as Moderators.” Electronic Commerce Research and Applications, 2024.

Michal Jacovi, Ido Guy, S. Kremer-Davidson, S. Porat, and Neta Aizenbud-Reshef. “The Perception of Others: Inferring Reputation from Social Media in the Enterprise.” Conference on Computer Supported Cooperative Work, 2014.

Michelle E. Daniels, and Freeman Wu. “No Comments (from You): Understanding the Interpersonal and Professional Consequences of Disabling Social Media Comments.” Journal of Marketing, 2024.

Nadia Steils, Annabel Martín, and Jean-François Toti. “Managing the Transparency Paradox Of Social-Media Infuencer Disclosures.” Journal of Advertising Research, 2022.

Osnat Roth-Cohen, Sigal Segev, and Yu Liu. “The Efect of Non-Celebrity Infuencers’ Perceived Authenticity on Social Media Advertising Outcomes.” International Journal of Advertising, 2024.

Rumen Pozharliev, D. Rossi, and Matteo De Angelis. “Consumers’ Self-Reported and Brain Responses to Advertising Post on Instagram: The Efect of Number of Followers and Argument Quality.” European Journal of Marketing, 2022.

S. Dzreke, and S. Dzreke. “The Infuencer Equity Equation: Analyzing Infuencer Marketing's Efect on Brand Equity via the Perspectives of Authenticity, Credibility, and Engagement.” International Journal For Multidisciplinary Research, 2025.

S. Jin, Ehri Ryu, and Aziz Muqaddam. “I Trust What She's #Endorsing on Instagram: Moderating Efects of Parasocial Interaction and Social Presence in Fashion Infuencer Marketing,” 2021.

S. Tong, Yingjie Zhang, Gordon Burtch, and Zhijie Lin. “Individual Sellers’ Social Media Participation and Sales Performance in Peer-to-Peer Marketplaces: Evidence From a Quasi-Natural Experiment.” Production and Operations Management, 2024.

Sándor Erdős. “The Efects of Community-Based Signals on Investment Decisions in Copy Trading.” Journal of Behavioral and Experimental Economics, 2022.

Sándor Erdős, Tamás K. Papp, and Zsófa Vörös. “The Efects of Social-Based Signals on Investment Decisions in Copy Trading.” Social Science Research Network, 2021.

Shunyuan Zhang, and Magie Cheng. “Impact of Brand Sponsorship on Infuencers-Audience Engagement: Evidence from Analyzing YouTube Videos.” Hawaii International Conference on System Sciences, 2023.

Suyash bansal. “The Impact of Social Media Infuencers on Consumer Behaviour.” International Scientifc Journal of Engineering and Management, 2024.

Tengteng Ma, Yingda Lu, Yuheng Hu, Xi Chen, and Yuxin Chen. “Content Creator Versus Brand Advertiser? The Efect of Inserting Advertisements in Videos on Infuencers Engagement.” Hawaii International Conference on System Sciences, 2023.

Van Canh Vu, Shasha Wang, Byron W. Keating, and E. Y. Chen. “Increasing Social Media Stickiness Through Parasocial Interaction and Infuencer Source Credibility.” Australasian Marketing Journal, 2024.

W. B. Elliott, Brian T. Gale, and S. Grant. “Navigating Through the Crowd: How Do Investors Assess Contributor Credibility and Make Investment Judgments on Social Media Platforms?” 2018.

Y. Z. C. Uğurhan, and İ. Yaşar. “Fenomene Ve Sponsorlu İçeriğe İlişkin Güvenin Marka Farkındalığı Üzerindeki Etkisinde Parasosyal İlişkinin Aracı Rolü: Youtube Özelinde Bir İnceleme.” İnsan Ve Toplum Bilimleri Araştırmaları Dergisi, 2021.

Yijia Cao, Yusuf Oc, Fang Wang, and Yanli Pei. “When We Are Alike: Homophily in Livestream Commerce.” Journal of Consumer Marketing, 2025.

Zeynep Karagür, Jan-Michael Becker, Kristina Klein, and Alexander Edeling. “How, Why, and When Disclosure Type Matters for Infuencer Marketing.” International Journal of Research in Marketing, 2021.

Zofa Saternus, P. Weber, and O. Hinz. “The Efects of Advertisement Disclosure on Heavy and Light Instagram Users.” Electronic Markets, 2022.

Appendix

Table 1.1

Anmol Bansal et al., 2025 No

M. Cheng et al., 2022 No

Michal Jacovi et al., 2014 Yes

Shunyuan Zhang et al., 2023 Yes

Osnat Roth-Cohen et al., 2024 No

M. Cheng et al., 2024 No

Rumen Pozharliev et al., 2022 Yes

S. Jin et al., 2021 No

M. G. Senali et al., 2024 No

G. L. Cascio Rizzo et al., 2023

Yes

Jihye Kim et al., 2024 No

Zeynep Karagür et al., 2021 Yes

Kyungsik Han et al., 2018 Yes

A. P. Tedjakusuma et al., 2025 No

Giuseppe Soda et al., 2024 No

Nadia Steils et al., 2022 No

Darlene Walsh et al., 2024 No

Tengteng Ma et al., 2023 Yes

D. C. Hugh et al., 2022 No

G. Wang et al., 2014 No

Sándor Erdős et al., 2022 No

Y. Z. C. Uğurhan et al., 2021 No

G. L. Cascio Rizzo et al., 2023a Yes

Komal Shamim et al., 2022 Yes

YouTube

IBM Connection (enterprise)

YouTube

Instagram

YouTube

Observational with DiD Beauty/style infuencers English-speaking

Observational 554 participants Global (IBM)

Observational with DiD 85,669 videos from 861 infuencers Global (English-speaking)

Observational survey 310 Instagram users United States

Observational with DiD Beauty/style infuencers English-speaking

Instagram Experimental (online + EEG) N=192 (online), N=112 (lab) European

Instagram Experimental 195 females Not specifed

Instagram

Instagram, TikTok

Not specifed

Instagram

Twitter

Live-streaming platform

Social trading platform

Not specifed

TikTok

Bilibili

Not specifed

SeekingAlpha, StockTwits

Mock-up social trading

YouTube

Instagram

Not specifed

Observational survey 416 individuals Not specifed

Mixed-methods 20,923 sponsored posts from 1,376 infuencers Not specifed

Survey-based 485 social media users United States

Mixed-methods (feld + experimental) 3,593 posts from 61 infuencers Germany

Mixed-methods 1,000 authors, 300 evaluators Not specifed

Observational survey 682 respondents Not specifed

Observational 28,000+ traders Not specifed

Mixed (observational + experimental) n=1,004 (Study 2) Not specifed

Not specifed Not specifed Not specifed

Observational with DiD 344,130+ videos from 5,000 infuencers China

Observational 281 followers Not specifed

Observational 9 years (SeekingAlpha), 4 years (StockTwits) Not specifed

Experimental Not specifed Not specifed

Observational Not specifed Not specifed

Mixed-methods 9,766 posts from 28 virtual infuencers US, Japan, UK

Observational survey

234 respondents Nepal

Yijia Cao et al., 2025 Yes

Van Canh Vu et al., 2024 Yes

Michelle E. Daniels et al., 2024 No

Estefania Ballester et al., 2025 No

Adrian Waltenrath et al., 2024 Yes

Bartosz Wilczyński et al., 2025 Yes

Edith Shalev et al., 2025 No

S. Dzreke et al., 2025 Yes

Zofa Saternus et al., 2022 Yes

Taobao Live

YouTube, Instagram, TikTok, Facebook

Twitter

Instagram

Instagram, YouTube, TikTok

Yelp, Goodreads, Twitter/X, Instagram

Instagram, TikTok, YouTube

Observational survey 313 respondents China

Observational survey N=417 (Study 1), N=249 (Study 2) Vietnam, United States

Mixed-methods Not specifed Not specifed

Observational survey 1,012 followers Not specifed

Observational 65,000+ posts from 239 macro-infuencers

Global (excluding DACH)

Cross-sectional survey 314 physiotherapy students Poland

Observational Four large datasets Not specifed

Mixed-methods N=172 (experimental) Not specifed

Instagram Experimental N=566 Germany

Ágnes Buvár et al., 2024 No Not specifed Experimental N=169 Not specifed

Suyash Bansal et al., 2024 No Not specifed

A. Rynarzewska et al., 2024 No

W. B. Elliott et al., 2018 No

M. Ghosh et al., 2025 No

S. Tong et al., 2024 No

Sándor Erdős et al., 2021 No

TikTok

Social media platforms

Facebook

Xianyu (Fishponds)

Mock-up social trading

Mixed-methods Not specifed Not specifed

Experimental Not specifed Not specifed

Observational 252 millennials Not specifed

Quasi-natural experiment 180,000+ transactions China

Experimental Not specifed Not specifed

Creator Trustworthiness Signals: What to Track Before You Partner

Proxies for Creator Trust

Abstract

Screening

TLDR: The Behavioral Proxy Approach to Creator Trustworthiness

1. The Paradigm Shift: From Metrics to Behavioral Proxies

2. The Universal Signal: Analyzing the Sponsored-Organic Performance Gap

3. Segmented Vetting: Micro-Influencer vs. Macro-Influencer Trust Signals

4. Environmental Conditioning: Platform Architecture and Content Format

5. The Disclosure Framework: Managing the Transparency Paradox

Strategic Implication

6. Implementation: The Behavioral Vetting Protocol

Our take

Characteristics of Included Studies

Engagement Metrics as Trust Proxies

Content Characteristics and Language Signals

Sponsored Content Disclosure as a Trust Signal

Creator-Audience Interaction Signals

Visual and Authenticity Markers

Expertise and Source Credibility Signals

Behavioral Consistency Signals

Synthesis

Influencer Size as a Critical Moderator

Platform Type Conditions Signal Effectiveness

Content Context and Commercial Intent

Audience Characteristics and Usage Intensity

Strongest Signal Proxies by Context

References