Redundant attention fails to replicate as sustained engagement: Information quality in early review cohorts and long‐run popularity of books

Abstract Early reviews on social reading platforms are an important clue for book buyers. However, research has so far focused on their volume and average rating rather than their informational content. To address this gap, we scraped nearly one million reviews of 25,000 books across five genres from lubimyczytac.pl, Poland's largest social reading platform. We modeled the long‐run accumulation of reviews attracted after a book's 30‐day initial launch window. Conditional on early review volume and book characteristics, reviews with greater semantic similarity and stronger solicitation cues predict less subsequent engagement. In contrast, early reviews with more linguistic perplexity and more diverse reviewer portfolios predict more engagement in the future. Furthermore, a text sentiment classifier we fine‐tuned to this specific domain added predictive signal beyond the raw star ratings, consistent with ratings being a coarsely discretized projection of the evaluative signal carried by review text. We frame these results within information quality theory: the informational profile of an early review cohort contributes to the prediction of sustained platform engagement beyond review volume and rating.