A significant fraction of massive stars in the Milky Way and other galaxies are located far from star clusters and star-forming regions. It is known that some of these stars are runaways, i.e. possess high space velocities (determined through the proper motion and/or radial velocity measurements), and therefore most likely were formed in embedded clusters and then ejected into the field because of dynamical few-body interactions or binary-supernova explosions. However, there exists a group of field O stars whose runaway status is difficult to prove via direct proper motion measurements (e.g. in the Magellanic Clouds) or whose (measured) low space velocities and/or young ages appear to be incompatible with their large separation from known star clusters. The existence of this group led some authors to believe that field O stars can form in situ. Since the question of whether or not O stars can form in isolation is of crucial importance for star formation theory, it is important to thoroughly test candidates of such stars in order to improve the theory. In this paper, we examine the runaway status of the best candidates for isolated formation of massive stars in the Milky Way and the Magellanic Clouds by searching for bow shocks around them, by using the new reduction of the Hipparcos data, and by searching for stellar systems from which they could originate within their lifetimes. We show that most of the known O stars thought to have formed in isolation are instead very likely runaways. We show also that the field must contain a population of O stars whose low space velocities and/or young ages are in apparent contradiction to the large separation of these stars from their parent clusters and/or the ages of these clusters. These stars (the descendants of runaway massive binaries) cannot be traced back to their parent clusters and therefore can be mistakenly considered as having formed in situ. We argue also that some field O stars could be detected in optical wavelengths only because they are runaways, while their cousins residing in the deeply embedded parent clusters might still remain totally obscured. The main conclusion of our study is that there is no significant evidence whatsoever in support of the in situ proposal on the origin of massive stars.