Many posit that recent increases in media polarization and specialization are a cause for recent political polarization, particularly in the United States. We develop a model of endogenous media bias, voter learning, and voting to test this mechanism. We find that in the old media environment (with no ability to target advertising, and hence no media specialization), convergence did not necessarily indicate truth-telling: if a sufficient portion of the population has a taste for ideological media, firms would bias towards moderation in order to avoid alienating potential viewers on either side. The ability to target advertising (and specialize) will generate media polarization, but at least one media firm reports the truth and will capture ideologically unbiased voters. Therefore, the total level of information transmitted weakly increases, and expected political polarization decreases. The model demonstrates why modern media polarization is an unlikely culprit for political polarization, and may actually attenuate divergence.