This paper provides evidence on the reliability of euro area real-time output gap estimates. A genuine real-time data set for the euro area is used, including vintages of several sets of euro area output gap estimates available from 1999 to 2006. It turns out that real-time estimates of the output gap are characterised by a high degree of uncertainty, much higher than that resulting from model and estimation uncertainty only. In particular, the evidence indicates that both the magnitude and the sign of the real-time estimates of the euro area output gap are very uncertain. The uncertainty is mostly due to parameter instability, while data revisions seem to play a minor role. To benchmark our results, we repeat the analysis for the US over the same sample. It turns out that US real time estimates are much more correlated with final estimates than for the euro area, data revisions play a larger role, but overall the unreliability in real time of the US output gap measures detected in earlier studies is confirmed in the more recent period. Moreover, despite some difference across output gap estimates and forecast horizons, the results point clearly to a lack of any usefulness of real-time output gap estimates for inflation forecasting both in the short term (one-quarter and one-year ahead) and the medium term (two-year and three-year ahead). By contrast, some evidence is provided indicating that several output gap estimates are useful to forecast real GDP growth, particularly in the short term, and some appear also useful in the medium run. No single output gap measure appears superior to all others in all respects.