Party reputations are an essential feature of party competition. Earlier scholarship identifies parties’ connections to social groups as an important constituent component of party reputations, and tends to see party reputations as stable in the short run. We challenge this view, arguing that group appeals, i.e. valenced references to social groups, can in fact lead to short-run changes in party-group linkages. We theorize that citizens keep running tallies of group appeals and frequently update perceptions of their group linkages in response to group appeals by party elites. We test the theory by examining how group linkages expressed in surveys track party elites’ group appeals in speeches in the UK House of Commons. To measure group appeals, we develop an approach that involves fine-tuning a BERT language model and use it to classify appeals by four parties to 13 groups across 500,000 sentences. The resulting data allows for tracking short-run changes in group appeals. We link the data to British Election Study panel surveys measuring citizens’ changing perceptions of each group-party linkage over several decades. In line with our running tally theory of group linkages, we find that group linkages robustly track parties’ use of group appeals. By our estimates, shifting just 10 appeals from neutral to positive over 3 months improves a perceived group-party linkage by 3 percentage points. The paper makes three contributions. First, we advance the measurement of group appeals, a key concept in party politics, and make an annotation model publicly available. Second, our analysis tests a key assumption in the group appeals literature that has never been examined outside survey experiments. Third, and more generally, we challenge a conventional view of party reputations as static, suggesting instead that party elites have considerable latitude to change party reputations in the short run.