It's notoriously difficult to track steps from the wrist, so it's not a surprise that comparing a wrist-based fitbit with an Apple watch will result in broadly similar counts, but comparing a body-worn fitbit such as the One or Zip will result in commonly lower count on on the watch than the fitbit.
One of the common issues with tracking from the wrist is that the accelerometers have to untangle much more complex motions, but at the same time, if the wearer is carrying something, or pushing a buggy for example, the movements are damped out and harder to accurately detect.
As an example, if I go on an hour-long walk in my neighborhood, my fitbit one and Apple watch will usually give readings within about 4-5% of each other, with the watch generally a little lower than the fitbit. If, on the other hand, I have a typical day which includes 45 minutes in the supermarket, some carrying of equipment, carrying my bag to and from work etc, the difference can be as much as 10,000 steps in a 30,000 step day, with the watch always lower.
The result is that I still use the fitbit, because it has always returned consistent results where wrist-based trackers never have. I don't bother with the fitbit site however, since I'm only monitoring my own activity for my own sake. Where the watch is very good is the inclusion of the activity rings, which quickly and graphically represent a snapshot of my day, and are highly motivating as a consequence. The actual numbers the activity app provides for movement and exercise aren't all that useful, though the stand reminders and 12 hour goal are very helpful.
One area where the watch seems to fare much better than any of the wrist-based trackers I have tried is that it appears far less prone to false-positives - counting arm movements, waving, clapping etc as steps when they are not.
