CAPTIONING POSTCARDS: VERBAL CUES ACROSS VISUAL THEMES FOR DMO INSTAGRAM ENGAGEMENT
Abstract
Visual storytelling increasingly shapes how destination marketing organizations (DMOs) engage audiences on social media. This study investigates how the linguistic and visual characteristics of DMO Instagram posts influence user engagement. Drawing on dual coding theory and processing fluency theory, we analyze 2,598 image-based posts from 33 Indian DMO accounts. Using deep learning-based image analysis and computer-assisted linguistic assessment, we examine how multimodal content characteristics shape user engagement. Image features are extracted with Google Cloud Vision API and clustered using Word2Vec with latent semantic analysis. Caption language is coded using the Linguistic Inquiry and Word Count framework. Negative binomial regression examines how these features affect likes and comments. The findings show that lexical complexity reduces both likes and comments, while first-person singular pronouns increase engagement, and collective “we” language decreases it. Positive emotion mainly increases likes, whereas tentative language stimulates comments, and certitude language decreases engagement. Past focus language generates different engagement outcomes depending on the visual context and engagement metric. The study contributes by showing that DMO Instagram engagement is theme-specific and metric-specific. It further reveals a fluency–dialogue trade-off in which cues that support passive liking may not equally stimulate active commenting.