We study the role of endogenous trust in amplifying ideological bias. Agents in our model seek to learn a sequence of states using information from sources whose accuracy is ex ante uncertain. Agents also receive noisy feedback about the states from direct observation. Arbitrarily small biases in the processing of this feedback can cause large ideological differences in the agents’ trust in information sources and their beliefs about the states, and may lead agents to become overconfident in their own judgment. These patterns can be similar whether agents see only ideologically aligned sources or see a diverse range of sources.