Do appeals that subtly invoke negative racial stereotypes shift whites’ political attitudes by harnessing their racial prejudice? Though widely cited in academic and popular discourse, prior work finds conflicting evidence for this “dogwhistle hypothesis.” Here we test the hypothesis in two experiments (total N = 1,797) in which white Americans’ racial attitudes were measured two weeks before they read political messages in which references to racial stereotypes were implicit, explicit, or not present at all. Our findings suggest that implicit racial appeals can harness racial resentment to influence policy views, though specifically among racially resentful white liberals. That dog-whistle effects would be concentrated among liberals was not predicted in advance, but this finding appears across two experiments testing effects of racial appeals in policy domains—welfare and gun control—that differ in the extent and ways they have been previously racialized. We also find evidence that the same group occasionally responded to explicit racial appeals even though these appeals were recognized as racially insensitive. We conclude by discussing implications for contemporary American politics, presenting representative survey data showing that racially resentful, white liberals were particularly likely to switch from voting for Barack Obama in 2012 to Donald Trump in 2016.