It's even more stark if you weight it by sampling online content rather than a sample of people online (who are mostly lurkers). Certain types of people tend to post a LOT, so a random sample of online content will be biased towards the "high posting frequency" type of person, who is probably not a normal average type of person.