We show that the implied cost of capital (ICC) can be a biased proxy for expected return, with the bias increasing in idiosyncratic volatility (IVOL) and expected growth. If not accounted for, this structured bias can lead to incorrect inferences about whether a firm characteristic represents a priced risk factor. Consistent with this concern, we find that commonly used disclosure measures, due to their inherent correlation with IVOL (but not with growth), exhibit a robust positive relation with ICC. We then propose and implement two simple bias-mitigation methods---controlling for model-implied bias drivers and subtracting an empirically estimated bias component from ICC---and find that the positive disclosure--ICC relation largely disappears once these adjustments are applied. Overall, the evidence underscores the importance of accounting for structured ICC bias when drawing ICC-based inferences about priced risk.