The increased availability of online knowledge has led to the design of several algorithms that solve a variety of tasks by harvesting the Semantic Web, i.e., by dynamically selecting and exploring a multitude of online ontologies. Our hypothesis is that the performance of such novel algorithms implicitly provides an insight into the quality of the used ontologies and thus opens the way to a task-based evaluation of the Semantic Web. We have investigated this hypothesis by studying the lessons learnt about online ontologies when used to solve three tasks: ontology matching, folksonomy enrichment, and word sense disambiguation. Our analysis leads to a suit of conclusions about the status of the Semantic Web, which highlight a number of strengths and weaknesses of the semantic information available online and complement the findings of other analysis of the Semantic Web landscape.