We show empirically that risk-neutral quantile skewness is priced differently depending on which portion of the risk-neutral distribution it is estimated from. Quantile skewness estimated from the tail (center) of the risk-neutral distribution is positively (negatively) related to future stock returns which explains contradicting results on the pricing of risk-neutral skewness in the literature. We argue that our results are consistent with ambiguity averse investors that avoid information from the extrapolated tail of the volatility surface. Quantile skewness is highly correlated with central skewness but more robust. We show in a simulation study that estimates of quantile skewness are accurate even if option prices only span a small domain, have large gaps between strikes, and are observed with noise.