Words of war: lexical salience in Sputnik Afrique articles and posting rhythms on X during the Russia–Ukraine war

Abstract Purpose This study examines lexical salience and qualitative framing in Sputnik Afrique articles, together with the temporal pattern of posting by its official French-language X account, during the first 158 days of Russia’s full-scale invasion of Ukraine. It distinguishes publisher output from reach, exposure, algorithmic effects, and disinformation prevalence, none of which can be inferred from public posting records or lexical recurrence alone. Design/methodology/approach The study uses a sequential mixed-methods single-case design. The lexical corpus contains 59,899 case-normalized French word-form tokens reconstructed from the titles, standfirsts, and article bodies of 150 Sputnik Afrique pages selected from 222 archived candidates in three event windows. A proportionally stratified, maximum-variation subset of 50 articles (21/10/19 by window) was systematically coded after full-text reading. Separately, 620 posts from @sputnik_afrique were summarized through monthly counts and posts per observed day across the 158-day period. Findings In the 59,899-token corpus, the selected recurrent forms Ukraine (472), Russie [Russia] (318), opération [operation] (169), and militaire [military] (161) were especially frequent. Six dominant frames were identified in the qualitative subset: military-institutional (11; 22 %), information-contestation (9; 18 %), territorial-security (8; 16 %), sanctions-economic (8; 16 %), civilian-humanitarian (7; 14 %), and diplomatic-institutional (7; 14 %). X data show intense publisher posting during the first five observation days and sustained activity thereafter, but do not measure reach, exposure, algorithmic amplification, or disinformation prevalence; claims were not independently fact-checked. Practical implications The study provides an auditable procedure for connecting aggregate lexical results with article-level framing while treating X counts strictly as records of publisher activity. It may support conflict-communication research and media monitoring without presenting posting volume as evidence of recommendation effects. Social implications By separating editorial framing, attributed source voices, and public posting activity, the study contributes to a more critical interpretation of wartime communication and platform metrics. Originality/value The article combines a fully specified 59,899-token corpus, rank-referenced lexical results, and structured article-level framing. This integration moves beyond frequency description while maintaining explicit evidentiary limits for claims about platforms and disinformation.