A new method is developed to estimate social influence in Internet communities that follow a specific developing news story. The technique stems from mean-field treatment of magnetic systems and provides a measure for community stability, such as the potential of a small perturbation to culminate in a phase-transition-like phenomenon. Three real cases of developing news stories from CNN news website are analyzed. Continuous dynamics of social influence together with time is observed together with a significant increase of social influence after the announcement of important information, such as the jury decision in a legal case. This work makes it possible to estimate the size of a group that can change the opinion of the entire population. We argue that Internet comments may predict the level of social response similar to a barometer that predicts the intensity of a coming storm in still calm environment.