Seventy years of parasocial research: a bibliometric analysis and science mapping, 1956–2025

Research using the term ‘parasocial’ has appeared since 1956 in media studies and psychology and, more recently, in influencer marketing. This study maps the development, intellectual structure and recent thematic direction of the 2174 journal works published from 1956 to 2025 that use the term ‘parasocial’ in its media sense. Using records from OpenAlex, it combines performance analysis with co-citation, bibliographic coupling, keyword mapping and topic analysis and tests the robustness of the main mapping results across alternative corpora, periods and network settings. Output increased from one article in 1956 to 521 works in 2025, and, by authors’ affiliations, the USA, Indonesia and China had the most works. Horton and Wohl’s 1956 article is cited by 63.1% of the works with a reference list. Under the main network settings, the most frequently cited references form two co-citation communities, which we interpret as a media-psychology foundation and an influencer-marketing stream; they differ in topic and in the age of their references, and the share of co-citation link weight between them increased across the three periods from 2014–2019 to 2023–2025. The keyword ‘Influencer marketing’ appears in 28.7% of the works published in 2023–2025; in that period the topic ‘Digital Marketing and Social Media’ exceeded ‘Media Influence and Health’, whereas across 2020–2025 the two were nearly equal (397 and 405 works). OpenAlex topic labels identify AI-related topics in 3.0% of the works from 2023–2025. A separate language-model-assisted classification of titles and stored abstracts classified 71 of these works (6.3%) as studying a parasocial bond with an artificial partner; the author examined every such case individually. Together, the widely shared cited reference, the rising co-citation between the two communities and the application of the concept to influencers and artificial agents offer a frame for comparing which conditions of a parasocial bond studies keep and which they change; how studies define or measure the phenomenon, and whether the concept applies to artificial agents, are questions that future content analyses and experiments could examine.