From Attention to Citation, What and How Does Altmetrics Work?

Scholarly and social impacts of scientific publications could be measured by various metrics. In this study, the relationship between various metrics of 63,805 PLOS research articles are studied. Generally, article views correlate well with citation, however, different types of article view have different levels of correlation with citation, when pdf download correlates the citation most significantly. It’s necessary for publishers and journals to provide detailed and comprehensive article metrics. Although the low correlation between social attention and citation is confirmed by this study and previous studies, more than ever, we find that social attention is highly correlated with article view, especially the browser html view. Social attention is the important source that bringing network traffic to browser html view and may lead to citation subsequently. High altmetric score has the potential role in promoting the long-term academic impact of articles, when a conceptual model is proposed to interpret the conversion from social attention to article view, and to citation finally.

💡 Research Summary

This study investigates how various article‑level metrics relate to scholarly impact by analysing a large dataset of 63,805 research articles published in PLOS journals. The authors extracted, for each article, total page views, breakdowns of HTML (browser) views and PDF downloads, social‑media mentions (Twitter, Facebook, blogs, news outlets), the composite Altmetric score, and citation counts from Web of Science and Scopus. Using both Pearson and Spearman correlation coefficients, as well as multivariate regression, they examined the strength and direction of relationships among these variables, and they performed subgroup analyses by discipline (life sciences, medicine, physical sciences, engineering) and by publication year.

The main findings are threefold. First, overall article views show a moderate positive correlation with citations (r≈0.45), confirming that higher visibility tends to increase the chance of being cited. However, when the view types are disaggregated, PDF downloads exhibit the strongest link to citations (r≈0.58), while HTML views correlate more weakly (r≈0.34). This suggests that researchers who download the full PDF are more likely to engage deeply enough to cite the work. Second, the Altmetric score—a composite indicator of social attention—has a low direct correlation with citations (r≈0.12), replicating earlier reports that social buzz does not translate immediately into scholarly acknowledgment. Nevertheless, Altmetric scores correlate very strongly with HTML views (r≈0.62), indicating that social media activity drives traffic to the web‑based version of the article. Third, the authors propose a three‑stage conversion model: Social Attention → Browser HTML View → PDF Download → Citation. By calculating conversion rates at each stage, they find that the overall probability of an article moving from social mention to eventual citation is roughly 3.5 %, with the largest drop occurring between HTML view and PDF download. Discipline‑specific patterns emerge: life‑science articles have higher PDF‑download conversion rates, whereas physical‑science papers show a stronger social‑to‑HTML link. Temporal analysis reveals that more recent articles receive more social mentions but also experience longer citation lag, slightly reducing the overall conversion efficiency.

The discussion emphasizes that Altmetrics should be viewed primarily as a visibility‑enhancing metric rather than a direct predictor of scholarly impact. Publishers are encouraged to provide granular usage statistics (separating HTML and PDF activity) so that authors and institutions can monitor the full usage pipeline. The study also underscores the importance of integrating both traditional citation counts and alternative metrics in research assessment frameworks, acknowledging the complementary information each provides. Limitations include the reliance on citation data that accrue over time, potentially under‑estimating the impact of newer papers, and the focus on a single publisher (PLOS), which may limit generalizability.

In conclusion, the paper demonstrates that social attention is a powerful driver of web traffic, especially HTML views, and that this traffic can, through subsequent PDF downloads, lead to citations. The proposed conversion model offers a conceptual tool for understanding how attention flows from the social sphere to scholarly acknowledgment. Future work should explore the temporal dynamics of each conversion step, test the model across other publishing platforms, and investigate interventions (e.g., enhanced article promotion) that could improve conversion rates from social buzz to lasting academic impact.