Visual digital intermediaries and global climate communication: Is climate change still a distant problem on YouTube?

PLOS ONE Visual digital intermediaries and global climate communication Table 2. (Continued) Lang Category #vid #cha #vie #com #lik %PB %Ice %PBI ja Science org. 3 1 43567 534 0 0.33 0.00 0.33 ja (Inter)gov. org. 2 2 8007 62 6 0.00 0.00 0.00 ja Pop. sci. groups 2 2 73106 2133 227 0.50 0.50 0.50 ja Civil society 1 1 1224 118 3 0.00 0.00 0.00 Portuguese Channel concentration: .11. Content: .00 (PB) .02 (PBI). Lang Category #vid #cha #vie #com #lik %PB %Ice %PBI pt (Inter)gov. org. 9 9 2289 133 1 0.00 0.00 0.00 pt Civil society 7 6 841 22 3 0.00 0.00 0.00 pt No description 4 4 1262 14 0 0.00 0.25 0.25 pt Pop. sci. groups 4 3 548 114 0 0.00 0.00 0.00 pt Online education 3 3 37395 1315 0 0.00 0.00 0.00 pt User-led ch. 3 3 13778 3401 135 0.00 0.00 0.00 pt Prof. media org. 1 1 2829 60 0 0.00 0.00 0.00 pt Religious org. 1 1 11263 301 34 0.00 0.00 0.00 pt Talk platform 1 1 1360 87 0 0.00 0.00 0.00 https://doi.org/10.1371/journal.pone.0318338.t002 assemblages underpinning the top-ranked results for ‘climate change’ on YouTube Search, and does this differ across region/language search contexts (RQ2)? Our findings suggest that the ranking assemblages, and the conditions of visibility that they entail, are fairly constant. On this mainstream climate keyword, ranking assemblages operate in largely similar ways across contexts, and this in turn translates into different prospects for fresh and locally rel- evant climate communication in particular contexts. We probe the platform, channel and audience sides of the ranking assemblage, and find that agencies on the respective sides

contribute. To start with the platform side, we do not observe the platform actively diversifying results or giving visibility to a greater number of channels. The API displays limited temporal diversi- fication: the top-20 results for region-language queries performed at different times are mostly the same. Fig 1 shows the fraction of shared videos between top-20 videos retrieved in queries repeated after one and two weeks (queries with less than 20 results not considered). Each data point represents a query for a region/language pair (e.g. Hindi in India, English in India), with the distribution ranging from 0 (no common results) to 1 (all same results). The median per- centage of common top-20 videos is high, respectively 85% and 75% after one and two weeks. Moreover, 75% of the queries returned at least 65% and 60% common results after one and two weeks. Since we got 400 results (our self-imposed limit) for almost all searches, we can rule out the possibility that there was no alternative supply. Meanwhile, the conditions of visibility for local actors are constrained. The API presents limited regional diversification. Searches in the same language return similar results for dif- ferent regions and continents, even when the language of the video itself is not in that lan- guage. Searches in neighbouring regions with different languages yield significantly different results. We tested on multiple collections, but here present results from a single data collec- tion. Fig 2 shows the overlap, i.e. fraction of shared videos, in top-20 results for all pairs of

country-language queries. Each cell shows a pair of queries, and the colour indicates over- lap, from no overlap (blue) to complete overlap (red). The clusters correspond to queries with the same language. The five largest are English (79 countries), French (35), Arabic (24), Span- ish (21), and Portuguese (10). Similar patterns remain when considering fewer and more results. Importantly, higher ranked results show the least diversification (see supplementary PLOS ONE https://doi.org/10.1371/journal.pone.0318338 April 8, 2025 14/ 23