You might think so, but some evidence suggests otherwise:
Much scholarship has suggested television has “personalized” voting behavior in American presidential elections by encouraging citizens to cast ballots on the basis of candidate image and personality. Though the assertion is common, little solid evidence exists that this is true. In this paper, I use National Election Studies data to examine whether voters are more concerned with candidates’ personal characteristics now than they were at the outset of the television era. Contrary to the popular view, I find voters no more likely today to mention candidate personality as a reason for their vote choice than they were in the 1950s and 1960s. Moreover, while personality affects voting behavior, its influence on candidate choice is not significantly larger than it was a half-century ago. The results are insensitive to a variety of model specifications and different measures of perceptions of candidate image. The findings are consistent with research on the resurgence of partisan voting in American elections.
That is from this paper by Danny Hayes. In fact, he finds that, when asked to list likes and dislikes about the candidates in an open-ended fashion, citizens are more likely to mention issues not personality.
Hayes’ account of the correlates of candidate choice dovetails with other studies (gated, JSTOR). The upshot: Americans are increasingly basing their choices on their own partisan leanings and not on more superficial aspects of the candidates.