The value of microblogging services (such as Twitter) and social networks (such as Facebook) in disseminating and discussing important events is currently under serious threat from automated or human contributors employed to distort information. While detecting coordinated attacks by their behaviour (e.g. different accounts posting the same images or links, fake profiles, etc.) has been already explored, here we look at detecting coordination in the content (words, phrases, sentences). We are proposing a metric capable of capturing the differences between organic and coordinated posts, which is based on the estimated probability of coincidentally repeating a word sequence. Our simulation results support our conjecture that only when the metric takes the context and the properties of the repeated sequence into consideration, it is capable of separating organic and coordinated content. We also demonstrate how those context-specific adjustments can be obtained using existing resources.