A Match Made in Language: Examining the Role of Linguistic Similarity in Adolescents’ Preference for Persuasive Health Messages in Social Media Contexts

Engaging adolescents with health topics on social media is notoriously challenging. A common strategy is message personalization, yet most efforts focus on personalizing <i>what</i> is said rather than <i>how</i> it is expressed. This is a missed opportunity, as youth-relevant health communication requires attention to both style and content. Digital communication environments enable such adaptation by making users' linguistic patterns visible and available for personalization. Yet little is known about how adolescents respond to health messages written in a similar linguistic style, or how such similarity is best operationalized. Addressing these gaps, this preregistered study tests whether adolescents prefer social media health messages that are more similar to their linguistic style and identifies which linguistic categories are most effective in eliciting positive responses. Using WhatsApp conversations donated by 191 Dutch adolescents (aged 13-15), we derived linguistic profiles and created Instagram-style health messages varying in linguistic similarity. Participants then evaluated 22 message pairs, each pair manipulating one linguistic category. Multilevel Bayesian analyses yielded inconclusive evidence for effects of linguistic similarity on message preference or personalization, and moderate evidence against effects on perceived effectiveness. Our findings point to potential trade-offs between experimental transparency and similarity strength, yielding important insights for refining future operationalizations. At the same time, adolescents consistently preferred positive and simple-worded messages, which gives general stylistic guidance for more engaging health communication on social media. Further, we discuss implications for AI-based style matching, including the use of algorithmic approaches and large language models.