Shouldn’t usage be the only criterion, whether enough people actually use the word? Surely dictionaries should reflect usage rather than attempt to create it.
Nope. That’d leave out zillions of academic, scientific, and profession-related words. Also, a word like “truthiness” or “santorum” can be used several million times, but if nearly all of those uses also refer to, mention, or quote their popularizer (Stephen Colbert and Dan Savage, respectively), then those millions of uses count for far less.