“One Face, Many Roles”: The Role of Cognitive Load and Authenticity in Driving Short-Form Video Ads ^†

Abstract

Short-form video platforms have shifted advertising from standalone, time-bounded spots to feed-embedded, swipeable stimuli, creating a high-velocity processing context that can penalize casting complexity. We ask whether a “one face, many roles” casting strategy (a single actor playing multiple characters) outperforms multi-actor executions, and why. A two-phase pretest (N = 3500) calibrated a realistic ceiling for “multi-actor” casts, then four experiments (total N = 4513) tested mechanisms, boundary conditions, and alternatives. Study 1 (online and offline replications) shows that single-actor ads lower cognitive load and boost account evaluations and purchase intention. Study 2, a field experiment, demonstrates that Need for Closure amplifies these gains via reduced cognitive load. Study 3 documents brand-type congruence: one actor performs better for entertaining/exciting brands, whereas multi-actor suits professional/competence-oriented brands. Study 4 rules out cost-frugality and sympathy using a budget cue and a sequential alternative path (perceived cost constraint → sympathy). Across studies, a chain mediation holds: single-actor casting reduces cognitive load, which elevates brand authenticity and increases purchase intention; a simple mediation links cognitive load to account evaluations. Effects are robust across settings and participant gender. We theorize short-form advertising as a context-embedded persuasion episode that connects information-processing efficiency to authenticity inferences, and we derive practical guidance for talent selection and script design in short-form campaigns.

1. Introduction

Short-form video platforms (e.g., TikTok, Douyin) have shifted advertising from discrete, stand-alone spots to in-feed, swipeable stimuli that compete with an endless stream of entertainment and user-generated content [1,2,3,4]. In these feeds, ads appear alongside organic videos and are skippable within seconds, placing a premium on immediate comprehension and rapid persuasion. Short-form video is therefore not merely “shorter television” but a distinct persuasion context marked by constrained processing time, rapid cue turnover, and algorithmic curation. Theoretical accounts of limited attentional resources are central in such contexts. Cognitive Load Theory (CLT) and the Limited Capacity Model of Motivated Mediated Message Processing (LC4MP) formalize how finite capacity is allocated to audiovisual material and how extraneous demands impair message processing [5,6,7,8,9].

This paper examines a creative choice pervasive on short-form platforms yet under-specified in advertising research: a “one face, many roles” casting strategy in which a single performer rapidly portrays multiple characters, contrasted with a conventional multi-actor execution. We propose that consolidating roles in one recognizable performer reduces extraneous cognitive load by minimizing identity tracking and role segmentation within seconds, thereby improving downstream persuasion relative to multi-actor formats in short-form settings [6,7,9]. We further argue that this processing relief facilitates authenticity inferences. When messages are easier to parse, people experience processing fluency—a metacognitive signal that shapes judgments of truth and genuineness [10,11,12]. In branding, perceived authenticity—the sense that a brand is true to its essence and values—is a robust antecedent of favorable evaluations and intentions [13,14]. Accordingly, we model a chain in which casting reduces cognitive load, which elevates authenticity, which in turn enhances purchase-relevant outcomes. In parallel, lower cognitive load should also improve evaluations of the creator’s account/channel that hosts the ad, yielding a simple mediation from casting to account evaluation via load.

Boundary conditions are expected at both the viewer and brand levels. Individuals higher in Need for Closure (NFC)—a preference for swift, unambiguous resolution—should benefit disproportionately from simplified identity tracking and thus exhibit larger gains from single-actor executions [15,16]. Brand-type fit should also matter: playful or “exciting” brands may find a rapid, multi-persona performance congruent with their identity, whereas competence-oriented brands may be better served by traditional multi-actor signals of professionalism [13,17,18]. For clarity, we conceptualize brand type as a boundary condition on the overall casting effect on account impression and purchase intention, rather than as a moderator of each mediation path.

We test these ideas in a multi-study program preceded by a large pretest to calibrate “multi-actor” complexity and followed by four experiments spanning online, offline, field, and text-based settings (total N = 4513). Across studies, we manipulate casting, measure cognitive load and authenticity with established scales, examine viewer- and brand-level moderators (NFC; brand type), and evaluate outcomes for both the ad host (account impression) and the advertised brand (purchase intention). We also rule out an alternative process account by testing whether perceived cost constraint and sympathy form a viable sequential pathway from casting to persuasion; they do not, and the single-actor advantage persists when budget cues are made explicit. Together, the evidence supports a dual-path account in which processing efficiency in micro-narratives is pivotal for short-form persuasion and authenticity emerges as a downstream evaluation rather than a mere stylistic cue.

The remainder of the article proceeds as follows. We review literature on cognitive load in mediated message processing and on brand authenticity to formalize hypotheses. We then detail the pretest and four experiments, present results, and conclude with theoretical and managerial implications, limitations, and avenues for future research.

2. Literature Review

Short-form video platforms (e.g., TikTok, Douyin, Instagram Reels, YouTube Shorts) have reshaped the attention ecology of advertising by embedding persuasive messages in algorithmically curated, swipeable feeds rather than in stand-alone ad pods [3,4,19]. Large-scale user-trace studies show that these feeds optimize for quick engagement and rapid skipping, constraining opportunities for elaboration and privileging designs that are instantly legible [20,21]. In this context, most short videos last only seconds and compete directly with adjacent entertainment content. Processing efficiency—especially the clarity of identity cues and the economy of visual elements—therefore becomes central to ad effectiveness [6,7,22]. The present review synthesizes evidence on cognitive capacity limits in mediated viewing and links those limits to judgmental outcomes that matter for short-form advertising.

2.1. Cognitive Load in Compressed, Feed-Embedded Viewing

Cognitive load theory (CLT) argues that limited working memory constrains learning and persuasion under time pressure [6,7]. In mediated environments, the Limited Capacity Model of Motivated Mediated Message Processing (LC4MP) further specifies how structural features (e.g., pacing, cuts, novelty) tax encoding and storage resources [8,23]. Within advertising, visual complexity reliably increases processing burden and reduces stopping power when complexity is not tightly managed [24,25]. These principles map directly onto short-form feeds: when messages must be parsed in a few seconds, executional choices that reduce the number of distinct elements to be tracked—such as consolidating multiple roles in a single on-screen face—should lower extraneous load and improve comprehension.

Face-processing research offers a mechanism for why who is on the screen matters. Humans rely on configural/holistic processing to individuate faces; tracking several new identities in rapid succession is effortful [26]. By having one actor re-costumed across roles, short-video ads can preserve narrative contrast while simplifying identity tracking, thereby conserving limited processing resources predicted by LC4MP and CLT [6,8].

Dual-process theories sharpen the boundary conditions. Under the Elaboration Likelihood Model and the Heuristic-Systematic Model, time pressure and divided attention push viewers toward heuristic processing; cues that enhance processing fluency thereby become more persuasive [27,28]. When a single performer links scenes, the ad may achieve higher perceptual fluency—facilitating liking and truth judgments produced by ease of processing [10,11]. At the individual-difference level, closure-/structure-seeking motives can moderate tolerance for ambiguity under speeded viewing [29], implying that the load-reducing advantage of one-actor execution may vary across audiences.

2.2. Casting Strategy and Narrative Coherence in Short Form

Classical advertising guidance—developed for longer, isolated media—often favors multiple actors to increase realism, role differentiation, and source credibility [17,18]. In short-form feeds; however, coherence under speed is paramount: viewers must infer “who is who” and “what is being offered” within seconds while anticipating the next swipe. Contemporary evidence from short-form ad research reinforces that content characteristics which streamline perception (e.g., a clear protagonist, simple transitions, consistent persona) enhance engagement and purchase responses [4,19,30]. Thus, while multi-actor casts can enrich plots in long form, consolidating roles in one performer may better fit the attentional constraints of feed-embedded persuasion by simplifying identity tracking and minimizing extraneous elements.

2.3. Authenticity as a Second Pathway from Execution to Persuasion

Beyond load reduction, short-video audiences often prize authenticity—signals of sincerity, integrity, continuity, and personal expression [13,14,31]. The construct is multifaceted: brand-authenticity work distinguishes continuity/heritage, credibility, integrity, and symbolism [13,14], while qualitative accounts emphasize indexical/iconic cues and the strategic downplaying of commercial motives [32]. In social media specifically, creators frequently deploy “calibrated amateurism”—a crafted yet unvarnished esthetic—to appear real and personally invested [33]. Influencer studies likewise show that passionate and transparent authenticity management fosters trust [34].

These perspectives imply that a single-actor/multiple-roles execution can signal authenticity in short videos: the visible “doing it all” performance conveys craft and personal involvement, aligning with platform-native expectations [13,14,33]. Processing fluency contributes here as well; fluently processed content often feels truer and more likable [10,11]. At the same time, authenticity judgments also reflect identity- and morality-linked assessments beyond mere ease of processing [17].

2.4. Integrating the Streams and Specifying the Gap

Taken together, the literature on cognitive load/LC4MP, dual-process/fluency, and brand/influencer authenticity converge on a common prediction: in short-form, feed-embedded contexts, a casting strategy that concentrates roles in a single performer should (a) minimize extraneous processing, (b) heighten perceived authenticity through platform-congruent, “handcrafted” cues, and (c) thereby improve downstream persuasion outcomes. Prior scholarship has not systematically linked casting format to these mechanisms under short-form constraints; much guidance is extrapolated from longer-form or isolated units [17,18]. Meanwhile, recent cross-platform work underscores the generality of short-form dynamics beyond TikTok to Reels and Shorts [21,22,30], yet it has not delineated how casting choices map onto load and authenticity in these feeds.

Accordingly, the present research positions cognitive load and perceived authenticity as dual mediators connecting casting to persuasion in short-video advertising, while recognizing boundary conditions shaped by viewers’ closure-/structure-seeking tendencies and brand context [13,14,29]. For clarity, brand type is conceptualized as a boundary condition on the overall casting effect on outcomes (account impression and purchase intention), rather than as a moderator of individual mediation paths.

3. Hypothesis Formulation and Research Framework

In long-form or standard television advertising, including cinematic spots and extended online videos, a larger and more diverse cast often fosters credibility, realism, and dramatic depth [18,31]. Those formats typically afford viewers minutes to learn who each person is and how their actions advance the plot. By contrast, TikTok-style short-form ads compress brand stories into 15–60 s, so every additional face must be encoded in a fraction of the time available; as a result, extra identities compete for limited processing resources and attention [6,7]. Cognitive Load Theory (CLT) formalizes this constraint by positing a severely bounded working memory; when an execution adds elements that must be attended to, individuated, and held in mind, comprehension and persuasion suffer [6,7]. The Limited Capacity Model of Motivated Mediated Message Processing (LC4MP) extends the same logic to audiovisual messages, arguing that structural features and additional cues draw on a single, finite pool of encoding and storage resources [8,9]. Advertising studies on visual complexity converge on this point: denser arrays with more focal elements impose higher processing burden and depress effectiveness unless complexity is actively managed [24,25].

Face perception research suggests a concrete mechanism through which casting multiplies cognitive requirements. Humans individuate faces largely via configural processing; once a single identity is encoded, subsequent sightings of that same face demand less effortful individuation [26]. A short-form execution in which one performer rapidly shifts costumes, postures, and vocal registers permits viewers to anchor each persona change to an already-built facial schema, rather than initiating fresh identification routines for unfamiliar actors. Taken together, these perspectives imply that a single-actor/multi-role format should compress the number of distinct identity cues that must be encoded, enhance perceptual fluency via repeated face-schema activation, and redirect spared capacity from “who’s who?” indexing to message comprehension. Accordingly,

H1. 

Relative to a multi-actor cast, deploying one performer in multiple roles will reduce viewers’ perceived cognitive load in short-video advertising.

If the one-actor format reduces extraneous load as posited in H1, then the immediate evaluative consequences should extend beyond comprehension to how viewers judge the source that “hosts” the message in a social-video feed. Dual-process theories hold that speeded, divided-attention contexts bias audiences toward heuristic processing, so cues that increase processing ease acquire greater persuasive weight [27,28]. When a message is fluently processed, people experience positive affect and a “feels-right” sensation that facilitates favorable judgments [10,11,12]. LC4MP further links higher load to adverse psychophysiological responses, which can spill over into less favorable content evaluations [9]. In social video, attitudes formed during processing often attach to the perceived communicator—frequently the creator’s account rather than a corporate channel—via source attribution and related parasocial mechanisms [35,36,37]. Lower load also frees attentional bandwidth for relational signals (e.g., eye contact, micro-timing of humor, narrative pacing) that support parasocial affinity, thereby improving account-level impressions [38,39]. Consequently, reducing cognitive effort should translate into more favorable evaluations of the account that curates the ad. Accordingly,

H2. 

Lower cognitive load will improve viewers’ evaluations of the short-video account that hosts the advertisement.

The same reduction in processing difficulty has implications for perceived brand authenticity, which we define as the extent to which a brand is seen as true to its essence and values [13,14]. A large literature shows that metacognitive ease shapes truth and genuineness judgments; information that is easier to process often feels more plausible and “real” [10,11]. In short-form advertising, a single-actor/multi-role execution decreases identity-tracking demands and reduces the need to re-segment roles across cuts, so parsing the narrative feels effortless. That sense of fluency is then (mis)attributed to brand sincerity and transparency rather than to executional simplicity per se, consistent with authenticity perceptions that are sensitive to cues of craft, integrity, and continuity [13,32,40]. Importantly, this reasoning does not claim that authenticity is merely fluency; rather, it posits a causal ordering under short-form constraints in which lower extraneous load enables clearer meaning construction, which in turn supports authenticity inferences when the performance reads as personally invested and “nothing-to-hide.” Accordingly,

H3. 

Lower cognitive load will enhance consumers’ perceptions of brand authenticity in short-video advertising.

Brand authenticity, once formed, is a robust antecedent of consequential outcomes, including identification, trust, and willingness to buy or pay a premium [13,14]. In short-form feeds, users make rapid go/no-go decisions about whether to act on an impression while scrolling; authenticity therefore functions as a high-leverage summary signal that converts fleeting attention into intention. If the one-actor format reduces load and thereby elevates authenticity perceptions as argued in H1 and H3, then higher authenticity should directly lift purchase intention. Accordingly,

H4. 

Higher perceived brand authenticity will increase consumers’ purchase intention toward the advertised brand.

The preceding arguments specify a processing-first logic in which casting structure primarily influences extraneous cognitive load, and this early constraint shapes evaluative judgments under short-form, feed-embedded viewing [6,7,8,9,28]. Building directly on that ordering, we now formalize the mediational and moderating relations without re-stating prior rationale.

Given that a single-actor/multi-role execution reduces identity-tracking demands, the resulting ease of processing should translate into more favorable impressions of the hosting account—a source-side evaluation formed in the stream where ads are encountered [35,36].

H5a (simple mediation).

Casting strategy → Cognitive load → Account evaluation.

Separately, the same reduction in cognitive effort should enable authenticity inferences about the brand and thereby lift purchase intention, consistent with authenticity’s role as a high-leverage evaluative signal in rapid-scroll contexts [13,14].

H5b (chain mediation).

Casting strategy → Cognitive load → Brand authenticity → Purchase intention.

The five hypotheses above jointly specify the sequential logic that undergirds our framework. Casting format affects extraneous cognitive load first, because individuating one face across roles is simpler than tracking multiple unfamiliar actors in seconds. Reduced load then pre-conditions evaluative processing in two ways: it improves account evaluations through fluency-driven affect and attention to relational cues (H2), and it enables authenticity inferences about the brand by making the message feel clear, candid, and personally delivered (H3). Authenticity, once elevated, translates into stronger purchase intention (H4). This ordering is consistent with CLT and LC4MP’s emphasis on capacity constraints as precursors to higher-order judgments, with dual-process models that assign greater weight to fluency under time pressure, and with authenticity research that links perceived sincerity to behavioral intentions [6,8,9,14,28]. It also coheres with evidence that adding novel sources within micro-narratives raises processing costs and attenuates persuasion—precisely the pattern we would expect if cognitive economy is the initiating step in the chain [4].

Individuals differ in their tolerance for ambiguity and desire for definitive answers. Need for Closure (NFC) captures this motive and predicts stronger preference for clear, quickly resolvable information structures [15,41]. In a seconds-long ad, a multi-actor cast proliferates identity cues and prolongs uncertainty about “who is who,” while a single-actor format collapses those cues into a consistent facial schema. Under ELM/HSM conditions where time pressure biases heuristic processing, high-NFC viewers are especially sensitive to ambiguity and thus benefit more from format choices that deliver immediate coherence [27,28]. Combining these perspectives with CLT/LC4MP, the predicted interaction is straightforward: the single-actor format should yield a larger reduction in cognitive load among high-NFC audiences because it resolves identity ambiguity at the point where resources are scarcest [6,8,9]. Accordingly,

H6. 

Need for Closure moderates the effect of casting on cognitive load: the load reduction produced by a single-actor/multi-role format (vs. multi-actor) is greater for high-NFC than for low-NFC individuals.

This moderation at the processing node implies downstream heterogeneity in the two mediated paths formalized above: stronger improvements in account evaluation (via load → source liking) and purchase intention (via load → authenticity → intention) for high-NFC segments, which prefer order and closure under temporal pressure.

Short-form viewers hold schematic expectations about how different brands “ought” to communicate. Exciting/entertaining brands are granted broader latitude for playful, persona-switching performances; serious/competence-oriented brands are expected to project reliability through more formal cues [13,14,17,18]. A single-actor format naturally foregrounds expressive craft and quick-fire role contrast—an esthetic that fits approach-oriented, fun brand identities—while a multi-actor layout more readily signals institutional competence and external validation in professional categories. Within our processing-first framework, this means that the same reduction in load can amplify authenticity when the performance feels brand-congruent (for entertaining brands), yet be offset by a schema violation that dampens authenticity when the performance feels flippant relative to category norms (for serious brands). Consequently, casting should interact with brand type on both consumer purchase intentions and account-level impressions. Accordingly,

H7. 

Brand type moderates the effect of a one-actor/multi-role casting strategy on (a) purchase intention and (b) positive account impressions: the strategy is more effective for entertaining/exciting brands than for serious/competence-oriented brands.

Figure 1 (Conceptual Framework) synthesizes Hypotheses 1–7 into a unified structural model. The diagram positions casting strategy (one-actor/multi-role vs. multi-actor) as the exogenous driver. Its primary effect proceeds through cognitive load—the central processing-efficiency mechanism identified in H1—and branches into two theoretically distinct consequence streams.

Collectively, the framework integrates information-processing theory with brand-personality and individual-difference perspectives, offering a parsimonious yet comprehensive account of how creative casting decisions in short-video advertising propagate through cognitive, affective, and contextual mechanisms to shape both brand-centric and account-centric outcomes.

4. Experiment

This study is divided into four studies, of which Study 1 includes two studies, 1a and 1b. The total number of samples is 4513, and all experiments were conducted in China.

4.1. Experiment 1

Experiment 1 examined whether casting structure influences short-form ad effectiveness by contrasting a single-actor/multi-role execution with a multi-actor/single-role execution. Because the literature provides no consensus threshold for what constitutes a “multi-actor” cast in short-form contexts, we conducted a two-phase pretest. In an online survey (N = 3500), over 92% of respondents judged more than six distinct actors as unacceptable for a short ad. In parallel, interviews with 15 experienced short-video creators indicated that six characters is the practical ceiling for maintaining engagement (based on view-through and like-rate heuristics). We therefore operationalized the multi-actor condition as involving up to six performers. To rule out contextual or gender confounds, Experiment 1 was implemented as two parallel studies: Experiment 1a (online; male lead) and Experiment 1b (offline; female lead), with random assignment to single-actor versus multi-actor conditions in both.

4.1.1. Experiment 1a (Online; Male Lead)

Design, participants, and stimuli. Participants were recruited via an online market-research panel and randomly assigned to one of two conditions (single actor vs. multiple actors). The stimulus was a 60 s short-form ad for a fictitious beverage brand, RefreshPlus, created to avoid prior familiarity. Both versions conveyed identical claims about taste, convenience, and health benefits; only casting differed. In the single-actor/multi-role version, one male performer enacted six characters in rapid succession via costume and mannerism shifts. In the multi-actor version, six distinct actors (three men, three women) each portrayed one character. Script, pacing, and visual style were held constant. After viewing, participants completed a brief unrelated reading task (~40 s) to mitigate recency, followed by the questionnaire and demographics. From 800 initial responses, standard data-quality screening (e.g., missing data, straight-lining/extreme patterns) yielded a final sample of N = 785. Descriptive statistics appear in Appendix A.

Measures. We used established multi-item scales with satisfactory reliability: Cognitive Load (CL), 4 items from [7], α = 0.802; Brand Authenticity (BA), 4 items from [13], α = 0.770; Purchase Intention (PI), 3 items from [42], α = 0.775; and Account Impression (AI), 4 items from [14], α = 0.848.

Manipulation check. Participants indicated perceived performer count on a 7-point item (“To what extent do you think the video was acted by one person?”; 1 = definitely many people, 7 = definitely one person). Scores were markedly higher in the single-actor condition (M = 6.41) than in the multi-actor condition (M = 2.43), t = 40.68, p < 0.001, confirming a successful manipulation.

Main effects (ANOVAs). Casting format significantly influenced all four outcomes:

Cognitive Load (CL): single-actor M = 2.51, SD = 0.82 vs. multi-actor M = 2.93, SD = 1.16; F(1, 783) = 33.64, p < 0.001, η² = 0.041. Brand Authenticity (BA): single-actor M = 5.54, SD = 0.74 vs. multi-actor M = 5.38, SD = 0.94; F(1, 783) = 7.49, p = 0.006, η² = 0.009. Purchase Intention (PI): single-actor M = 5.31, SD = 0.96 vs. multi-actor M = 4.82, SD = 1.17; F(1, 783) = 41.53, p < 0.001, η² = 0.050. Account Impression (AI): single-actor M = 5.52, SD = 0.98 vs. multi-actor M = 5.00, SD = 1.23; F(1, 783) = 42.56, p < 0.001, η² = 0.052.

These results support H1–H4.

Mediation tests (PROCESS). We conducted two mediation analyses with 5000 bootstrap samples (Hayes’ PROCESS).

H5a (simple mediation, CL → AI; Model 4). Casting predicted CL, which in turn predicted AI. The indirect effect was significant (effect = 0.2232, 95% CI [0.1451, 0.3048]). The direct effect of casting on AI remained significant (effect = 0.2947, 95% CI [0.1554, 0.4340]), indicating partial mediation and supporting H5a.

H5b (chain mediation, CL → BA → PI; Model 6). The CL-only path was significant (0.1693, 95% CI [0.1024, 0.2404]). The BA-only path was not significant (–0.0018, 95% CI [−0.0246, 0.0228]). Crucially, the sequential path CL → BA was significant (0.0396, 95% CI [0.0198, 0.0652]). The total indirect effect was significant (0.2212, 95% CI [0.1399, 0.3092]), consistent with cognitive load reduction as the primary mediator and authenticity as a downstream channel in the causal chain, supporting H5b.

Gender robustness checks. Because the independent variable is a human casting strategy and responses are formed by human viewers, gender is a theoretically plausible influence (performer gender, viewer gender, and their fit may shape impressions). In line with reviewers’ request to examine gender effects, we conducted two 2 × 2 ANOVAs using viewer gender (male vs. female) and casting (single-actor vs. multi-actor) as factors. With PI as the dependent variable, casting remained significant, gender showed no significant main effect, and the Casting × Gender interaction was not significant. The same pattern held for AI: a robust main effect of casting, no gender main effect, and no interaction. These findings rule out viewer-gender confounding for Experiment 1a (full results in

Table 1. The effect of gender on the dependent variable in Experiment 1.

Two-Factor ANOVA on PI (Experiment 1a)				Two-Factor ANOVA on AI (Experiment 1a)
Variable	F(1781)	p	η2	Variable	F(1781)	p	η2
CS	41.955	<0.001	0.051	CS	42.64	<0.001	0.052
Gender	0.693	0.405	0.001	Gender	0.941	0.332	0.001
CS × gender	0.456	0.500	0.001	CS × gender	0.374	0.541	0.000
Two-factor ANOVA on PI (Experiment 1b)				Two-factor ANOVA on AI (Experiment 1b)
variable	F(1782)	p	η2	Variable	F(1782)	p	η2
CS	39.104	<0.001	0.048	CS	44.385	<0.001	0.054
Gender	0.013	0.910	0.000	Gender	0.006	0.940	0.000
CS × gender	1.160	0.282	0.001	CS × gender	0.001	0.982	0.000

Source: This table was made by the authors. Note: CS represents the independent variable Casting Strategy.

).

Summary. Using a controlled 60 s stimulus with matched scripts and pacing, Experiment 1a shows that consolidating roles in a single performer reduces cognitive load and improves both brand- and account-level responses. The sequential mediation from Casting → CL → BA → PI is supported, and the simple mediation from Casting → CL → AI is also confirmed. Importantly, these effects are not contingent on viewer gender, while retaining external validity for a broad adult sample.

4.1.2. Experiment 1b (Offline; Female Lead)

Design, participants, and stimuli. Experiment 1b replicated Experiment 1a in an offline environment to rule out potential confounds of digital testing. The materials were identical to 1a except for the lead performer’s gender: in the single-actor/multi-role condition, a female performer enacted all six roles; in the multi-actor condition, six distinct performers (three women, three men) each played one role. Narrative, visual style, pacing, and role content were held constant across conditions. We distributed 800 paper questionnaires in public areas; after standard quality checks, the final sample comprised N = 786 valid responses. Participants were randomly assigned to the single-actor or multi-actor condition with approximately equal cell sizes, and no imbalances in age, gender, or education were observed across conditions. Descriptive statistics appear in Table 1.

Procedure and measures. All participants viewed a 60 s short-form ad for the fictitious beverage brand RefreshPlus (message claims about taste, convenience, and health benefits were matched across conditions). Immediately after viewing, participants completed a brief filler task (reading an unrelated paragraph) to reduce recency effects, followed by the questionnaire and demographics. The same validated multi-item scales were used as in 1a, with satisfactory internal consistency: Cognitive Load (CL), 4 items adapted from [7], α = 0.793; Brand Authenticity (BA), 4 items from [13], α = 0.733; Purchase Intention (PI), 3 items based on [42], α = 0.772; and Account Impression (AI), 4 items from [14], α = 0.804. Demographics (age, gender, education) were collected at the end of the survey.

Manipulation check. Perceived performer count was assessed with the item “To what extent was the ad performed by one person?” (1 = definitely many people, 7 = definitely one person). Ratings were higher in the single-actor condition (M = 5.894) than in the multi-actor condition (M = 2.340), t = 28.741, p < 0.001, confirming that viewers clearly recognized the casting difference.

Main effects (ANOVAs). One-way ANOVAs with casting condition as the independent variable showed:

Cognitive Load (CL): single-actor M = 2.641 vs. multi-actor M = 2.956; F(1, 784) = 17.617, p < 0.001, η² = 0.022. Brand Authenticity (BA): no reliable difference; F(1, 784) = 2.366, p > 0.05, η² = 0.003 (single-actor M = 5.460; multi-actor M = 5.363). Purchase Intention (PI): single-actor M = 5.291 vs. multi-actor M = 4.782; F(1, 784) = 38.649, p < 0.001, η² = 0.047. Account Impression (AI): single-actor M = 5.480 vs. multi-actor M = 4.938; F(1, 784) = 44.593, p < 0.001, η² = 0.054.

These results replicate the Experiment 1a pattern: the single-actor format lowered CL and increased PI and AI, while BA did not differ significantly by condition.

Mediation tests (PROCESS). Bootstrap mediation analyses (Hayes’ PROCESS, 5000 samples) examined the prespecified paths.

H5a (simple mediation, Casting → CL → AI; Model 4). The indirect effect via CL was significant (effect = 0.1354, 95% CI [0.0691, 0.2053]). The direct effect of casting on AI also remained significant (0.4070, 95% CI [0.2591, 0.5551]), indicating partial mediation.

H5b (chain mediation, Casting → CL → BA → PI; Model 6). Specific indirect effects were: Casting → CL → PI, 0.0943 (95% CI [0.0472, 0.1494])—significant; Casting → BA → PI, −0.0077 (95% CI [−0.0381, 0.0241])—not significant; Casting → CL → BA → PI, 0.0341 (95% CI [0.0148, 0.0601])—significant. The total indirect effect was 0.1208 (95% CI [0.0513, 0.1961]); the direct effect on PI remained significant (0.3886, 95% CI [0.2406, 0.5367]). These findings again support sequential mediation in which cognitive-load reduction is primary and authenticity functions as a downstream channel.

Gender robustness checks. Because both the manipulation (casting by human performers) and responses (human viewers) are person-centered, we assessed robustness to viewer gender using two 2 × 2 ANOVAs with casting (single-actor vs. multi-actor) and viewer gender (male vs. female) as factors. For PI, casting showed a positive main effect, while neither the main effect of gender nor the Casting × Gender interaction was significant. The same pattern held for AI: a positive main effect of casting, with no significant gender main effect and no interaction. Full results are reported in

Table 2. The effect of gender on the dependent variable in Experiment 2.

Three-Factor ANOVA on PI (Experiment 2)				Three-Factor ANOVA on AI (Experiment 2)
Variable	F(1973)	p	η2	Variable	F(1973)	p	η2
CS	52.336	<0.001	0.051	CS	35.477	<0.001	0.035
PER	0.387	0.534	0.000	PER	1.231	0.268	0.001
Gender	0.507	0.477	0.001	Gender	0.092	0.762	0.000
CS × PER	37.132	<0.001	0.037	CS × PER	24.592	<0.001	0.025
CS × Gender	1.766	0.184	0.002	CS × Gender	0.867	0.352	0.001
PER × Gender	1.361	0.244	0.001	PER × Gender	0.387	0.534	0.000
CS × PER × Gender	3.567	0.059	0.004	CS × PER × Gender	0.005	0.945	0.000

Source: This table was made by the authors. Note: CS represents the independent variable Casting Strategy.

.

Summary. Replicating Experiment 1a in an offline context with a female lead, Experiment 1b shows that consolidating roles in one performer reduces cognitive load and improves purchase intention and account impressions relative to a multi-actor execution, with brand authenticity unchanged across conditions. The mediation pattern—Casting → CL → BA → PI (chain) and Casting → CL → AI (simple)—is again supported, and the results are robust to viewer gender.

4.2. Experiment 2

Building on Experiment 1, Experiment 2 investigates Need for Closure (NFC) as a moderator of the effect of casting format (single-actor/multi-role vs. multi-actor) on consumer responses to short-form advertising. We test whether individuals who prefer quick, unambiguous resolutions respond more favorably to the single-actor format under field conditions. To assess robustness outside the lab, we implemented a between-subjects field experiment in a neighborhood supermarket.

4.2.1. Method

We used a 2 (Casting Format: single-actor/multi-role vs. multi-actor) × 2 (NFC: high vs. low) between-subjects design in collaboration with a local small supermarket. We initially recruited 1000 shoppers; after excluding incomplete responses and attention-check failures, the final sample was N = 981. Cell sizes were: single-actor/high-NFC (n = 243), single-actor/low-NFC (n = 245), multi-actor/high-NFC (n = 247), and multi-actor/low-NFC (n = 246). Descriptive statistics are reported in Appendix A.

Participants were randomly assigned to read one of two detailed textual ads for the supermarket. The single-actor script asked readers to imagine one male performer rapidly switching costumes and voices to portray six customer characters; the multi-actor script described six different performers (four men, two women) each playing one customer role. Scripts were equally vivid and realistic, and no images or video were shown to avoid appearance confounds.

NFC was manipulated via scenario priming (high-pressure vs. relaxed decision context) and verified with a 4-item NFC scale (α = 0.785). After reading the scripts, participants completed the same outcomes as in Experiments 1a/1b using validated multi-item scales: Purchase Intention (3 items; [42]; α = 0.814), Account Impression (4 items; [14]; α = 0.829), Cognitive Load (4 items; [7]; α = 0.741), and Brand Authenticity (4 items; [13]; α = 0.817).

4.2.2. Results and Discussion

Manipulation checks. Perceived performer count differed strongly by casting: single-actor M = 6.162 (n = 488) vs. multi-actor M = 2.992 (n = 493), t = 30.85, p < 0.001. NFC manipulation also succeeded: high-NFC M = 4.991 (n = 490) vs. low-NFC M = 3.886 (n = 491), t = 14.194, p < 0.001.

Cognitive load. A 2 × 2 ANOVA revealed a significant main effect of casting, F(1, 977) = 158.298, p < 0.001, η² = 0.139, with lower CL in the single-actor format. The Casting × NFC interaction was significant, F(1, 977) = 26.629, p < 0.001, η² = 0.027. Simple-effects analyses showed that within the single-actor condition, high-NFC participants reported lower CL (M = 2.816) than low-NFC participants (M = 3.107), F(1, 977) = 16.681, p < 0.001; within the multi-actor condition, high-NFC participants reported higher CL (M = 3.949) than low-NFC participants (M = 3.581), F(1, 977) = 10.352, p < 0.01.

Purchase intention and account impression. Casting affected PI, F(1, 977) = 49.473, p < 0.001, η² = 0.048, with higher means for single-actor (M = 5.456) than multi-actor (M = 4.924). AI showed a similar pattern, F(1, 977) = 34.549, p < 0.001, η² = 0.034, with single-actor M = 5.622 vs. multi-actor M = 5.221.

Moderated mediation. Using Hayes’ PROCESS, we tested the conditional indirect effects on PI (Model 83). The path Casting → Cognitive Load → PI was significant and stronger for high-NFC: Low-NFC indirect effect = 0.0931 (95% CI [0.0477, 0.1470]); High-NFC indirect effect = 0.2225 (95% CI [0.1198, 0.3336]); Index of moderated mediation = 0.1294 (95% CI [0.0589, 0.2205]).

Adding BA as a second mediator (Casting → CL → BA → PI) yielded: Low-NFC indirect effect = 0.0374 (95% CI [0.0198, 0.0605]); High-NFC indirect effect = 0.0895 (95% CI [0.0567, 0.1281]); Index = 0.0521 (95% CI [0.0290, 0.0814]).

The BA-only path (Casting → BA → PI) was not significant (effect = −0.0368, 95% CI [−0.0811, 0.0051]). Figure 2 depicts the moderating pattern.

For account impression (Model 7), the conditional indirect effect Casting → CL → AI was 0.0902 (95% CI [0.0453, 0.1420]) for low-NFC and 0.2156 (95% CI [0.1184, 0.3146]) for high-NFC; index of moderated mediation = 0.1254 (95% CI [0.0580, 0.2026]). The direct effect of casting on AI was 0.2480 (95% CI [0.1063, 0.3987]). Please see Figure 2.

Gender robustness checks. To assess potential gender influences, we estimated three-factor ANOVAs with Casting (single- vs. multi-actor) × NFC (high vs. low) × Gender (male vs. female) for PI and AI. In both models, Casting remained positive and NFC moderated the effect; all main or interaction terms involving gender were nonsignificant. Full results appear in Table 2.

Summary. Under field conditions, the single-actor/multi-role format reduced cognitive load and increased purchase intention and account impression relative to a multi-actor execution. The Casting → CL → BA → PI sequential path and the Casting → CL → AI simple path were both supported, with stronger conditional indirect effects among high-NFC shoppers. These findings extend the earlier experiments and confirm H6 in a real-world setting.

4.3. Experiment 3

Experiments 1–2 documented an overall advantage for the single-performer/multi-role format. Experiment 3 tests H7 by examining whether brand type moderates this casting effect. We contrast an entertaining brand (a trendy lifestyle headphone, EchoPop) with a professional brand (a digital tax-filing service, Finosure), following the logic that these brand contexts differ along Aaker’s brand-personality dimensions. We assess whether the one-actor/multiple-roles format is especially effective for entertaining brands, whereas a multi-actor execution is more suitable for professional brands.

4.3.1. Method

Design and stimuli. We employed a 2 (Casting Format: one-actor × multiple-roles vs. multi-actor) × 2 (Brand Type: entertaining vs. professional) between-subjects design using two fictitious brands: EchoPop (playful lifestyle headphones) and Finosure (no-nonsense digital tax filing). To eliminate potential confounds from facial appearance, voice quality, or acting style, all materials were presented in text. Specifically, this experiment was conducted on the online Credamo platform. Descriptive statistics appear in Appendix A.

Participants first read a brand-information sheet (logo, slogan, 90-word description) written either in a lively, slang-infused tone (EchoPop) or a formal, trust-focused tone (Finosure). Next, they read a richly worded storyboard for a 60 s social media ad for the same brand:

One-actor condition. A single male spokesperson “rapid-fire” switches among six roles—college gamer, commuter, fitness buff, music producer, student, and budget-conscious parent—with brief bracketed costume cues; the text explicitly stated that all roles are played by the same person.

Multi-actor condition. The parallel script listed the same six roles, explicitly stating that six different actors (four women, two men) appear, each delivering one segment. Role order, product claims, and scene descriptions were identical across conditions; only the casting attribution differed.

Both scripts highlighted identical product benefits: flair and energy for the entertaining brand versus clarity and reliability for the professional brand. Because materials were entirely textual, any observed effect isolates casting structure rather than performer looks, voice, or on-screen charisma.

Participants. From 1000 initial respondents, standard data-quality screening yielded N = 978 valid cases. Cell sizes were: One-Actor/Entertaining (n = 246), One-Actor/Professional (n = 243), Multi-Actor/Entertaining (n = 245), and Multi-Actor/Professional (n = 244). Random assignment produced no significant demographic differences across cells (e.g., age, gender), and cell sizes were roughly equal, ensuring counterbalancing and adequate power.

Measures. We used standard multi-item scales with good reliability. Purchase Intention (PI): 3 items adapted from [42], α = 0.796. Account Impression (AI): 4 items from [14], α = 0.842. Manipulation checks included: (a) Brand-type check (“This brand feels more fun and exciting,” 7-point scale; reverse-coded in the professional condition), and (b) Casting check (“How many different actors appeared in the video?”, open-ended; higher numbers indicate more actors). Cronbach’s alphas above 0.79 indicate acceptable internal consistency.

Manipulation checks. The casting manipulation was salient: the mean reported number of actors was 6.092 in the one-actor condition versus 2.967 in the multi-actor condition, t = 27.27, p < 0.001. The brand-type manipulation also succeeded: on the “fun/exciting” item, the entertaining brand was rated significantly more fun/exciting than the professional brand (M__Entertaining = 2.849 vs. M__Professional = 4.458), t = −13.72, p < 0.001; note that the professional brand’s higher numeric value reflects reverse coding for the serious/”professional” version.

4.3.2. Results and Discussion

Purchase Intention (PI). A two-way ANOVA with factors Casting (one-actor vs. multi-actor) and Brand Type (entertaining vs. professional) revealed a significant Casting × Brand Type interaction, F(1, 974) = 87.99, p < 0.001, η² = 0.083, indicating that casting effectiveness depends on brand context. Simple-effects tests showed a crossover:

Entertaining brand: one-actor M = 5.771 > multi-actor M = 5.061, F(1, 974) = 55.52, p < 0.001, η² = 0.054.

Professional brand: multi-actor M = 5.451 > one-actor M = 4.894, F(1, 974) = 37.70, p < 0.001, η² = 0.034.

Account Impression (AI). The ANOVA likewise yielded a significant Casting × Brand Type interaction, F(1, 974) = 62.29, p < 0.001, η² = 0.060. Simple-effects tests paralleled PI:

Entertaining brand: one-actor M = 5.584 > multi-actor M = 5.103, F(1, 974) = 22.33, p < 0.001, η² = 0.022.

Professional brand: multi-actor M = 5.372 > one-actor M = 4.714, F(1, 974) = 41.39, p < 0.001, η² = 0.041.

Numeric summary. PI—Entertaining: One-Actor 5.77 vs. Multi-Actor 5.06; Professional: Multi-Actor 5.45 vs. One-Actor 4.89. AI—Entertaining: One-Actor 5.58 vs. Multi-Actor 5.10; Professional: Multi-Actor 5.37 vs. One-Actor 4.71. All simple effects were p < 0.001. These results support H7: brand type moderates the effect of casting format. Figure 3 visualizes the pattern.

Cross-experiment note on cast composition. The pattern is consistent independent of actor-gender composition across studies. Experiments 1–2 varied gender balance yet showed advantages for the one-actor format; Experiment 3 used a female-dominant multi-actor description while replicating the predicted brand-by-casting crossover (one-actor best for the entertaining brand; multi-actor best for the professional brand). This consistency indicates that casting format per se, interacting with brand context, drives persuasion.

Gender robustness checks. We further estimated three-factor ANOVAs with Casting (one- vs. multi-actor) × Brand Type (entertaining vs. professional) × Gender (male vs. female) for PI and AI. In both models, Casting and the Casting × Brand Type interaction remained significant; all main or interaction terms involving gender were nonsignificant. Full results appear in

Table 3. The effect of gender on the dependent variable in Experiment 3.

Three-Factor ANOVA on PI (Experiment 3)				Three-Factor ANOVA on AI (Experiment 3)
Variable	F(1970)	p	η2	Variable	F(1970)	p	η2
CS	1.252	0.263	0.001	CS	1.574	0.210	0.002
BT	12.732	<0.001	0.013	BT	16.913	<0.001	0.017
Gender	0.340	0.560	0.000	Gender	2.168	0.141	0.002
CS × BT	87.550	<0.001	0.083	CS × BT	62.403	<0.001	0.060
CS × Gender	0.571	0.450	0.001	CS × Gender	2.382	0.123	0.002
BT × Gender	0.482	0.488	0.000	BT × Gender	0.858	0.355	0.001
CS × BT × Gender	1.166	0.280	0.001	CS × BT × Gender	2.590	0.108	0.003

Source: This table was made by the authors. Note: CS represents the independent variable Casting Strategy.

.

Summary. Experiment 3 demonstrates a brand-contextual boundary condition for casting: one-actor/multi-role outperforms multi-actor for an entertaining brand, whereas multi-actor outperforms one-actor for a professional brand. Manipulations were successful, effects were substantial for both PI and AI, and the pattern held under gender robustness checks. These findings align with H7, indicating that short-form casting choices should be matched to brand type to maximize persuasive impact.

4.4. Experiment 4

Short-form video ads often rely on condensed storytelling and multiple character roles. Experiments 1–3 showed that a single-actor/multiple-roles execution yields stronger consumer responses than a multi-actor execution. A plausible alternative account is cost attribution: viewers might infer that the single-actor format reflects budget frugality, which could elicit sympathy and, in turn, improve evaluations, independent of any cognitive benefit. Experiment 4 tests this alternative pathway by manipulating budget cues and measuring Perceived Cost Constraint (PC) and Sympathy (SP) as potential sequential mediators. Please see Figure 4 for the main theoretical framework of the alternative explanation of Experiment 4. Descriptive statistics are reported in Appendix A.

4.4.1. Method

Design and participants. We used a 2 (Casting Format: One-Actor vs. Multi-Actor) × 2 (Budget Cue: High Budget vs. No Cue) between-subjects design. 1000 shoppers were approached at a medium-sized bookstore café; after exclusions, the final sample was N = 983. Cell sizes were: One-Actor/No-Cue (n = 247), One-Actor/High-Budget (n = 246), Multi-Actor/No-Cue (n = 245), Multi-Actor/High-Budget (n = 245). Participants were randomly assigned to conditions.

Stimuli. All participants read a brand-information sheet for a fictional wellness app (BalanceOne) and then a text-only ad scenario (to avoid appearance/voice confounds). In the One-Actor condition, a single female spokesperson was described as performing six roles (stressed worker, yoga instructor, freelancer, insomniac student, overworked parent, nutritionist). In the Multi-Actor condition, six distinct actors (three men, three women) enacted the same six roles. Narrative and claims were identical across casting formats; only the number of actors differed.

Budget cue manipulation. In the High-Budget condition, text stated that BalanceOne “invested heavily in production,” hired a creative agency, and deliberately chose the casting strategy for storytelling reasons (i.e., not to save money). In the No-Cue condition, budget information was omitted. The manipulation aimed to shift perceived budget constraint without changing the storyline.

Measures. After reading, participants completed: Purchase Intention (PI): 3 items [42], α = 0.810. Account Impression (AI): 4 items [14], α = 0.847. Perceived Cost Constraint (PC): 4 items [43], α = 0.750. Sympathy (SP) toward the spokesperson: 4 items [44], α = 0.793. All scales used 7-point Likert response formats. Standard manipulation checks assessed perceived actor count and perceived budget level.

4.4.2. Results and Discussion

Manipulation checks. Both manipulations succeeded. Perceived performer count differed by casting: One-Actor M = 6.079 (7-point “one vs. many”) vs. Multi-Actor M = 2.955, t = 29.232, p < 0.001. Perceived budget level was higher under the High-Budget cue: M = 4.965 vs. No-Cue M = 3.016, t = 17.690, p < 0.001.

Purchase Intention (PI). A 2 × 2 ANOVA with Casting Format (One-Actor vs. Multi-Actor) and Budget Cue (High Budget vs. No Cue) showed a strong main effect of Casting, F(1, 979) = 105.968, p < 0.001, η² = 0.098: participants in the One-Actor groups (N = 493) reported higher PI (M = 5.485, SD = 0.949) than those in the Multi-Actor groups (N = 490; M = 4.757, SD = 1.249). The main effect of Budget Cue was not significant, F(1, 979) = 1.449, p > 0.05, η² = 0.001, and the Casting × Budget interaction was not significant, F(1, 979) = 0.004, p > 0.05, η² = 0.000. These results indicate that explicitly communicating a high production budget did not diminish the One-Actor advantage.

Account Impression (AI). A parallel 2 × 2 ANOVA likewise revealed a significant main effect of Casting, F(1, 979) = 52.812, p < 0.001, η² = 0.051: the One-Actor groups (N = 493) yielded higher AI (M = 5.655, SD = 0.871) than the Multi-Actor groups (N = 490; M = 5.515, SD = 1.252). The main effect of Budget Cue was not significant, F(1, 979) = 0.390, p > 0.05, η² = 0.000, and so was the Casting × Budget interaction, F(1, 979) = 0.117, p > 0.05, η² = 0.000. Figure 5 depicts the outcomes.

Sequential mediation (PC → SP). To assess whether PC and SP explain the Casting effect, we estimated Hayes’ PROCESS Model 6 with 5000 bootstrap samples, coding Casting as 0 = Multi-Actor, 1 = One-Actor and specifying PC (M1) → SP (M2) as sequential mediators.

Outcome: PI. X→PC→PI: indirect effect = −0.0017, 95% CI [−0.0132, 0.0070]. X→SP→PI: indirect effect = 0.0034, 95% CI [−0.0034, 0.0154]. X→PC→SP→PI: indirect effect = 0.0005, 95% CI [−0.0022, 0.0038]. Total indirect: 0.0022, 95% CI [−0.0088, 0.0158]. Direct effect (X→PI): 0.7262, 95% CI [0.5878, 0.8651], p < 0.001.

Thus, neither PC nor SP, alone or in sequence, accounted for the PI effect; the direct Casting effect remained large and significant.

Outcome: AI. X→PC→AI: indirect effect = −0.0019, 95% CI [−0.0100, 0.0143]. X→SP→AI: indirect effect = 0.0028, 95% CI [−0.0134, 0.0066]. X→PC→SP→AI: indirect effect = 0.0004, 95% CI [−0.0019, 0.0034]. Total indirect: 0.0013, 95% CI [−0.0100, 0.0143]. Direct effect (X→AI): 0.4988, 95% CI [0.3637, 0.6338], p < 0.001.

Again, PC and SP did not transmit the Casting effect on AI.

Additional regressions. Regressions predicting PI and AI from Casting (0/1), Budget Cue (0/1), PC, and SP confirmed that Casting was the only significant predictor: PI: β = 0.712, t = 7.100, p < 0.001; AI: β = 0.466, t = 4.769, p < 0.001; Budget Cue, PC, SP were ns. Please see

Table 4. The effect of gender on the dependent variable in Experiment 4.

Three-Factor ANOVA on PI (Experiment 4)				Three-Factor ANOVA on AI (Experiment 4)
Variable	F(1975)	p	η2	Variable	F(1975)	p	η2
CS	107.329	<0.001	0.099	CS	53.145	<0.001	0.052
Budget	1.457	0.228	0.001	Budget	0.350	0.554	0.000
Gender	2.085	0.149	0.002	Gender	0.116	0.734	0.000
CS × Budget	0.004	0.949	0.000	CS × Budget	0.156	0.693	0.000
CS × Gender	2.414	0.121	0.002	CS × Gender	0.925	0.336	0.001
Budget × Gender	0.573	0.449	0.001	Budget × Gender	0.887	0.346	0.001
CS × Budget × Gender	0.017	0.897	0.000	CS × Budget × Gender	1.152	0.283	0.001
OLS Regression on PI (Experiment 4)				OLS Regression on AI (Experiment 4)
Variable	Effect	p	Standard Error	Variable	Effect	p	Standard Error
CS	0.729	<0.001	0.071	CS	0.499	<0.001	0.069
Budget	−0.101	0.167	0.073	Budget	0.032	0.655	0.071
Gender	0.032	0.282	0.030	Gender	0.031	0.294	0.029
PC	−0.056	0.102	0.034	PC	−0.050	0.139	0.033
SY	−0.103	0.144	0.071	SY	−0.025	0.717	0.069

Source: This table was made by the authors. Note: CS represents the independent variable Casting Strategy.

for details.

Budget cue on PC (check). As expected, the High-Budget cue reduced perceived cost constraint: M__High-Budget = 3.111 (n = 491) vs. M__No-Cue = 3.595 (n = 492), t = −7.022, p < 0.001. This validates the intended shift in PC but, critically, does not translate into PI or AI differences.

Gender robustness checks. We ran three-factor ANOVAs with Casting (One vs. Multi) × Budget (High vs. No Cue) × Gender (male vs. female) for PI and AI. In both models, Casting remained significant; all main or interaction terms involving gender were nonsignificant. At the same time, this article also eliminated the possible influence of gender on the results of this experimental study through regression analysis. Complete results are provided in Table 4.

Summary. The One-Actor advantage on PI and AI persists under a high-budget cue and is not mediated by Perceived Cost Constraint or Sympathy, either alone or in sequence (PC → SP). The budget manipulation successfully lowered PC but did not affect outcomes, and neither PC nor SP accounted for the Casting effect. Together with prior studies, these results indicate that the persuasive advantage of consolidating roles in a single performer does not hinge on frugality-based attributions; rather, it is consistent with a processing-efficiency account established earlier.

5. General Discussion

Across four complementary experiments spanning online video (1a), offline replication (1b), a field setting (2), and text-only scripts that strip away surface cues (3–4), we find convergent evidence that a single-performer/multiple-roles casting strategy improves short-video ad effectiveness relative to conventional multi-actor executions. Mechanistically, consolidating roles in one face reduces viewers’ extraneous processing demands (H1), and this efficiency spills over to more favorable account evaluations (H2) and, via higher perceived authenticity, to stronger purchase intentions (H3–H4). Mediation analyses consistently show a chain of influence—Casting → Cognitive Load → Brand Authenticity → Purchase Intention—with cognitive load as the primary driver and authenticity as a downstream conduit (H5a–H5b). These effects are not uniform: they intensify for individuals high in Need for Closure (Experiment 2; H6) and reverse by brand type such that one-actor casting benefits “entertaining/exciting” brands but can disadvantage “serious/competence-oriented” brands (Experiment 3; H7). Finally, Experiment 4 rules out an alternative explanation that attributes the one-actor advantage to perceived frugality or sympathy: a high-budget cue neither attenuates the effect nor renders sympathy or perceived cost constraint diagnostic in mediation, reinforcing a processing-based account.

Taken together, the results suggest that casting—often treated as a creative executional choice—has theoretically predictable consequences in feed-based, swipeable environments. The same narrative economy that makes single-performer ads easy to follow (lower extraneous load) also makes them feel more candid and less contrived (higher authenticity), thereby translating scarce attentional windows into persuasion. The pattern generalizes across delivery modes (video vs. text vignettes), settings (lab-like vs. field), and lead gender (male/female leads in 1a/1b), while remaining sensitive to audience motives (Need for Closure) and brand-schema expectations. Conceptually, the findings position short-form advertising not as “shorter TV,” but as a context-embedded persuasion episode in which capacity constraints and heuristic inferences interact with brand-fit considerations to shape outcomes.

5.1. Theoretical Contributions

5.1.1. Reframing Casting in Feed-Based Persuasion Through Capacity Limits

Classic advertising wisdom holds that more characters enrich realism and credibility [17,18]. Our results qualify this view for short-form, feed-embedded contexts: when exposure windows are measured in seconds, every additional face competes for limited working memory resources, elevating extraneous load [6,7]. By anchoring multiple personas to a single encoded face, one-actor ads reduce individuation demands and preserve resources for message comprehension—an implication consistent with limited-capacity perspectives on mediated processing [9]. We thus extend traditional casting guidance to an environment where processing efficiency rather than scene richness becomes pivotal.

5.1.2. Establishing a Chain Mechanism That Links Cognitive Load to Authenticity-Based Persuasion

The study integrates cognitive load research with authenticity scholarship by demonstrating a sequential pathway: reductions in extraneous load produce metacognitive ease (processing fluency), which audiences heuristically map onto “this seems genuine,” elevating brand authenticity and purchase intention [11,12,13,14]. Importantly, the authenticity step is downstream of cognitive load rather than an independent channel, clarifying why the BA-only indirect effect is unreliably significant whereas the CL → BA link is robust across studies. This mechanism specifies how micro-narrative design can manufacture authenticity inferences without invoking artisanal production cues, broadening authenticity theory beyond craftsmanship and provenance to include cognitive-experiential fit in fast media.

5.1.3. Embedding the Account in Dual-Process Persuasion Frameworks

By lowering cognitive burden, one-actor casting shapes both the ability component of elaboration and the heuristic inferences that follow when ability/motivation are constrained—core ideas in dual-process theories such as the Elaboration Likelihood Model and the Heuristic–Systematic Model [27,28,45]. Our moderation by Need for Closure (Experiment 2) formalizes how a chronic desire for decisiveness amplifies the value of executional simplicity: high-NFC viewers benefit most from identity streamlining, showing greater load reductions and stronger downstream effects. The framework thereby connects a creative lever (casting) to psychologically grounded route selection in short-video persuasion.

5.1.4. Specifying Boundary Conditions via Brand–Schema Congruity

Experiment 3 demonstrates that brand personality moderates the casting effect: for “entertaining/exciting” brands, quick persona switching is congruent, reinforcing authenticity and evaluations; for “serious/competence-oriented” brands, the same device can appear flippant, muting authenticity and lowering persuasion. This brand-fit dependence complements person-level moderation and links casting to schema-congruity accounts in branding [13,17,18], offering a principled basis for when less is more and when more is warranted.

5.1.5. Ruling out Frugality-Driven Alternatives to Isolate a Processing Account

An alternative view is that single-performer ads look “low-budget,” eliciting sympathy or thrift admiration. Experiment 4 manipulates budget salience and measures perceived cost constraint and sympathy in a chain (PC → Sympathy), finding that neither the direct nor sequential paths explain outcomes; the one-actor advantage persists even under explicit high-budget cues. This strengthens the theoretical claim that the observed benefits are not artifacts of cost inferences but flow through capacity management and authenticity-as-fluency.

Collectively, these contributions elevate casting from a purely esthetic choice to a micro-architecture of processing in short-video ecosystems (e.g., TikTok/Douyin, Instagram Reels, YouTube Shorts). They also provide a transferable template for theorizing other creative levers—editing density, captioning, or pacing—through their impacts on cognitive economy and heuristic signaling in fragmented media.

5.2. Practical Contributions

Short-form, feed-embedded environments reward clarity on the first pass. Our evidence shows that a single performer enacting multiple roles is often the most efficient way to deliver that clarity. Viewers identify one face while tracking several personas, which reduces processing friction and helps the narrative “click” within seconds. In practical terms, creative teams planning sub-60 s placements on TikTok, Douyin, Instagram Reels, or YouTube Shorts can treat a one-actor/multi-role execution as a strong baseline for high-attention moments, especially when the objective is rapid comprehension and immediate persuasion.

Creative fit with brand identity remains essential. For brands that prize excitement, playfulness, or youthful energy, one-actor role-switching naturally signals spontaneity and creative confidence. The same device can feel off-tone for competence-centric categories (e.g., finance, healthcare), where multiple presenters or domain experts better cue professionalism and trust. Practitioners should therefore align casting with brand-personality frameworks rather than defaulting to a single recipe [17,18]. When the brand is “fun,” a versatile performer can heighten perceived authenticity by making the message feel personal and handcrafted. When the brand is “serious,” a multi-actor arrangement can legitimize expertise without overloading viewers.

Audience segmentation tactics can further unlock value. People differ in their tolerance for ambiguity and preference for fast closure. Marketers can anticipate stronger gains from one-actor formats among consumers who favor decisiveness and unambiguous messaging. That profile can be approximated with platform analytics (e.g., low dwell on complex ads, higher completion when narratives are linear) and then targeted or A/B tested. Creative variants can be tuned to the same logic: one-actor spots may adopt straight-line “set-up → payoff” storytelling, whereas multi-actor variants can be reserved for segments that enjoy layered narratives without experiencing overload.

Talent selection and briefing should conserve attention for meaning rather than identity parsing. Casting a performer with demonstrated range—vocal, facial, and physical—helps role transitions remain legible at a glance. Wardrobe and prop changes should be minimal but diagnostic; one clear cue per role is often sufficient. On-screen text can label roles sparingly to prevent clutter. When appropriate, a concise end card or caption can make the creative choice transparent (e.g., “All roles portrayed by [Name] to illustrate [Benefit]”). Such transparency reinforces intentionality without inviting budget attributions and can enhance perceived authenticity by foregrounding craft rather than thrift [14,31].

Measurement and optimization should reflect the mechanisms identified in our studies. Pre-launch tests can probe perceived mental effort and message clarity alongside standard brand metrics. In-platform signals such as early-scroll rates, view-through, and short rewinds provide pragmatic proxies for processing ease in fast feeds. For campaigns that pair brand-building with response objectives, decision rules can privilege one-actor variants when attention curves are steep (e.g., mobile peak hours or crowded feeds) and switch to multi-actor variants when credibility cues are paramount (e.g., regulatory updates or service explanations). Influencer collaborations deserve the same discipline. Creators known for sketch or impersonation styles are natural partners for one-actor executions, provided the tone stays on-brand. Creators with professional gravitas are better suited to multi-voice formats that spotlight expertise.

Production signaling should be managed deliberately. Our data indicate that explicitly communicating “high budget” does not weaken the one-actor advantage, suggesting that budget inferences are not the driver of effectiveness. Lightly signaling the creative rationale—through behind-the-scenes clips, pinned comments, or press notes—can preempt misinterpretations and amplify the sense of design intentionality. The goal is not to justify cost but to curate how audiences read the craft choices that make short-video storytelling legible, memorable, and persuasive.

Taken together, these recommendations emphasize fit and focus. Short-video advertising is not a compressed version of television; it is a different persuasion episode with different constraints. Teams that match casting to brand personality, tune variants to audience processing styles, and measure success with attention-aware diagnostics can harness the one-actor/multi-role strategy to cut through clutter while preserving credibility where it matters [14,17,18,31].

5.3. Limitations and Future Directions

The present program of studies was designed to isolate casting format as the focal executional lever in short-form video advertising, and that focus necessarily imposes scope conditions. One limitation concerns brand familiarity and equity. To avoid carryover from prior attitudes, we relied on fictitious or unfamiliar brands in several studies. In applied settings, preexisting equity, spokesperson fit, and celebrity reputations may either amplify or dampen the effects reported here by shifting baseline credibility or altering authenticity attributions. Follow-up work could embed the same casting manipulations within well-known brands or ongoing creator–brand relationships to estimate how equity and endorsement histories shape the net persuasive return of a single-actor approach.

A second limitation involves message genre and goal diversity. Our scenarios emphasized lighthearted, informational-to-persuasive micro-narratives, which are common on feed-based platforms but not exhaustive. Short videos can also be testimonial, instructive, or somber, and those genres may recruit different cues to credibility than the ones highlighted here. It remains an open question whether a one-actor execution supports delicate topics (e.g., safety, finance, health) without diluting perceived competence, or whether an ensemble cast conveys gravitas more effectively. Mapping the boundary conditions across genres—documentary-style, testimonial, episodic storytelling, or high-action formats—would help translate the present mechanisms to a fuller portfolio of creative objectives.

Beyond cost-frugality and sympathy, real-world responses can also be shaped by perceived entertainment value and performer charisma. Our designs mitigate these influences by using text-only storyboards with constant scripts (Studies 2–4) and matched pacing and claims in video stimuli (Studies 1a/1b), which isolate casting structure from performance style. Nevertheless, future work should explicitly measure and, where feasible, manipulate entertainment value and performer charisma—or include them as covariates or fixed effects—to further bound their impact in feed-embedded settings.

A third limitation relates to additional person- and context-level moderators beyond those tested. We theorized and found moderation by Need for Closure and by brand type, but heterogeneity may also arise from platform familiarity, advertising involvement, sensation seeking, device constraints, or viewing context (e.g., in transit versus at home). These factors plausibly alter tolerance for cognitive burden and the threshold at which “many roles” create confusion, and they could be examined through targeted sampling or platform analytics. Although our samples spanned a broad adult age range, short-video consumption skews young on many platforms; future research can probe whether the casting effects replicate or shift in magnitude in Gen-Z–heavy cohorts and in older segments that consume short video regularly.

A further consideration is ecological validity in a highly audiovisual medium. Text-only stimuli in Studies 2–4 were a deliberate choice to isolate identity-cue density from facial appearance, voice quality, and editing, but the trade-off is reduced sensory fidelity relative to real feeds. Converging designs that combine controlled casting manipulations with richer production—alongside objective traces such as view-through rates, dwell time, or scroll latency—would strengthen external validity without reintroducing confounds. Physiological and process-tracing measures (e.g., eye tracking, secondary-task reaction time) could also triangulate the cognitive-load account while preserving the feed-embedded experience.

Finally, the present studies used brief, validated multi-item reports for cognitive load, authenticity, account impression, and purchase intention within single sessions. This common structure facilitates comparability across experiments but does not exhaust the space of measurement strategies. Future work might supplement self-reports with behavioral choice tasks, follow-ups on delayed recall, or platform-native outcomes (e.g., click-through or follow intent realized in subsequent behaviors). Together, these extensions would clarify where the one-actor advantage holds most strongly and how it scales with brand equity, creative genre, audience composition, and real-world viewing conditions, while preserving the theoretical core that casting-induced efficiency in identity processing cascades into authenticity and persuasion in short-form feeds.