Timely identification of endangered populations is vital to save them from extirpation. Here we tested whether six commonly used early-warning metrics are useful predictors of impending extirpation in laboratory rotifer (Brachionus calyciflorus) populations. To this end, we cultured nine rotifer clones in a constant laboratory environment, in which the rotifer populations were known to grow well, and in a deteriorating environment, in which the populations eventually perished. We monitored population densities in both environments until the populations in the deteriorating environment had gone extinct. We then used the population-density time series to compute the early-warning metrics and the temporal trends in these metrics. We found true positives (i.e. correct signals) in only two metrics, the standard deviation and the coefficient of variation, but the standard deviation also generated a false positive. Moreover, the signal produced by the coefficient of variation appeared when the populations in the deteriorating environment were about to cross the critical threshold and began to decline. As such, it cannot be regarded as an early-warning signal. Together, these findings support the growing evidence that density-based generic early-warning metrics—against their intended use—might not be universally suited to identify populations that are about to collapse.