Send message to Authors

Paper short abstract

This analysis of AI co-authorship across 185 countries (2015–2024) reveals expanding geographic participation but persistent hierarchy. New centres emerge yet concentrate activity within subregions, suggesting that expanded collaboration does not automatically strengthen research capability.

Paper long abstract

The spread of artificial intelligence has intensified concern over unequal access to digital infrastructure, compute and skills. Less attention has been paid to the international relationships through which countries acquire the capacity to produce and evaluate AI knowledge. This article examines international AI co-authorship between 2015 and 2024 across five research fields, 185 countries and 19 subregions. Using a leading-partner network, it investigates whether expanding participation has redistributed research activity and strengthened countries’ capacity to sustain collaboration through several consequential relationships. The global AI research system became more extensive and more polycentric. India rose sharply in global activity, Saudi Arabia entered the leading country tier, and Singapore emerged as an important gateway in South-eastern Asia. The United Arab Emirates, South Korea, Taiwan and Australia also gained prominence, especially in frontier fields. Nevertheless, activity, continuity and partner alternatives remained highly uneven. New centres widened the geography of the upper tier while frequently concentrating activity within their own subregions. The article characterises this pattern as hierarchical integration: incorporation into a common research system expands, but the institutional and relational conditions for benefiting from that incorporation remain unequal. The concept helps distinguish simple entry from the durable conversion of collaboration into technological capability. For development policy, the relevant question is therefore not only how many countries participate, but whether collaboration strengthens continuing research activity, regional and domestic institutions, consequential partner alternatives and a greater role for local actors in defining research priorities and evaluation.