An Experimental Study of Human and Artificial Intelligence Collaboration in TikTok Advertising: Effects on Audience Perception and Engagement

Electronic Journal of Education, Social Economics and Technology Vol. 6, No. 2, (2025), pp. 1~14, Article ID: 1165 ISSN 2723-6250 (online) DOI: https://doi.org/10.33122/ejeset.v6i2.1165 Research Article An Experimental Study of Human and Artificial Intelligence Collaboration in TikTok Advertising: Effects on Audience Perception and Engagement Nadhif Muhammad Kasyfan, Btari Mariska Purwaamijaya*, Muhammad Dzikri Ar Ridlo Department of Digital Business, Universitas Pendidikan Indonesia, Tasikmalaya, Indonesia, 46115 *Corresponding Author: | Phone: +6285722237114 ABSTRACT The use of generative artificial intelligence, or GenAI in short form video advertising, continues to grow due to its efficiency and scalability. However, questions remain regarding audience acceptance, perceived authenticity, and the effectiveness of AI-generated advertising on social media platforms such as TikTok. This study examines how different creative configurations, namely Human Only, AI Only, and Human-AI Collaboration, influence audience perception and engagement in TikTok advertising. This study employed an experimental mixed methods approach. Quantitatively, two sequential A B split tests were conducted using TikTok Ads Manager to compare Human Only versus AI Only advertising and Human Only versus Human-AI Collaboration advertising. Audience engagement was measured using Completion Rate as the primary indicator and six-second view rate as a secondary indicator. Qualitatively, in-depth interviews with active TikTok users were conducted to examine audience processing mechanisms based on the Elaboration Likelihood Model. The results show that Human Only advertising achieved higher completion rates and early engagement than AI Only advertising. Furthermore, Human-AI Collaboration generated the highest engagement compared to Human-Only advertising. Qualitative findings indicate that human involvement strengthens perceived authenticity and trust, while AI supports visual structure that sustains attention and message elaboration. In conclusion, Human-AI Collaboration represents the most effective and socially acceptable approach to short-form video advertising, with implications for digital advertising strategy, ethical AI use, and the sustainable integration of generative technologies in social media communication. Keywords: Artificial Intelligence Generated Content (AIGC); Completion Rate; Elaboration Likelihood Model; Generative AI; Mixed-Methods; Human-AI Collaboration; TikTok Advertising 1. INTRODUCTION The digital advertising industry is undergoing a major transformation driven by the need for content that is fast, relevant, and engaging, especially on social media platforms (Faruk et al., 2021; Sholih & Almas Ashar, 2025). A key driver is Generative AI (Gen AI), an evolution beyond rule-based systems (Narrow AI) toward models that can creatively produce new content from large-scale data (Russell & Norvig, 2021; Dwivedi et al., 2023). Gen AI is now widely adopted by global brands such as Cadbury, Coca-Cola, Microsoft, McDonald’s, and Samsung (AdWeek, 2024; Forbes, 2024), because it can accelerate production, reduce costs, and generate high-quality content (Kang & Lou, 2022; Hartmann et al., 2025). The technology is reshaping not only content creation but also consumer interactions and decision-making processes (Grewal et al., 2025; Mogaji & Jain, 2024). This aligns with broader evidence that digital transformation technologies reshape online and social marketing practice(Dyah Kusumastuti et al., 2025). Industry leaders recognize the scale of this shift, with 70% of CEOs believing Gen AI will significantly change their businesses within the next three years (PwC, 2024). Despite its promise, large-scale adoption of Gen AI brings challenges. AI-generated advertising frequently raises ethical concerns and questions about audience acceptance (Wu & Wen, 2021). Projections indicate that AI could replace more than 200,000 jobs within the next 3 to 5 years (Bloomberg, as cited by Exploding Topics, 2025) and prompt about 14% of the global workforce approximately 375 million people to switch careers by 2030 (McKinsey, as cited by Exploding Topics, 2025).These concerns create tension between human creators and Gen AI (Anantrasirichai & Bull, 2022) and fuel AI Anxiety, a fear that human creativity will be displaced. Studies report that 36% of respondents experience this anxiety, more than 40% of consumers have lower trust in AI-based ads (Haupt et al., 2025; Kim et al., 2025), and 52% of young people aged 18 to 24 worry about its impact on their career prospects (BMG, as cited by Exploding Topics, 2024). Page 1 of 14 Kasyfan et al. Electronic Journal of Education, Social Economic and Technology, Vol. 6, No. 2, (2025), pp.1~14, Article ID: 1165 In practice, two dominant approaches are used to produce Gen AI advertising content: (1) AI-Only (pure AIGC), in which the entire creative process and core assets are automatically generated by AI while humans provide minimal intervention such as initial prompting or basic post-production like cut-to-cut editing, and (2) Human–AI Collaboration, in which AI acts as a creative assistant that augments rather than replaces human creators (Gao et al., 2023; Haupt et al., 2025; Zhang et al., 2022; Sowa & Przegalinska, 2025). The collaborative approach is generally seen as more acceptable because it preserves message credibility (Baek et al., 2024; Haupt et al., 2025). Even so, challenges remain. For example, transparency about content provenance can lower audience preference even when human involvement is present (Lefkeli et al., 2024). Despite the ongoing debate and these challenges, a significant gap persists in the academic literature. Mogaji & Jain (2024) underscore the need for continued research on consumer behavior in the Gen AI era. Prior studies have focused heavily on text-based or static visual content, while empirical work on Gen AI in short-form video advertising remains limited (Kang & Lou, 2022; Chung et al., 2025). Madathil (2025) compared AI-Only and Human–AI Collaboration using a multi-method approach, although the focus was on YouTube. This leaves a gap on TikTok, a platform with a distinctive engagement algorithm. This study addresses the gap through a mixed-methods approach grounded in the Elaboration Likelihood Model (ELM) (Cacioppo et al., 1986; Petty et al., 1983). ELM posits two routes to attitude formation. The central route relies on deep cognitive processing of argument quality. The peripheral route relies on affective cues. In this study, the quantitative experiment focuses on Completion rate as the primary metric and 6-second view rate as a secondary metric. Other indicators, such as likes, comments, and CTR, are recorded as exploratory due to their low frequencies. Qualitative in-depth interviews are then conducted to probe participants’ cognitive and emotional processes. Together, these methods aim to contribute a unique perspec (...truncated)