It largerly depends on how you are marking the route in Google Earth and how you actually ride it. There are a couple of factors that introduce small diferencies when measuring distances and their effects are multiplied over distance, so the longer the route, the bigger the difference.
The straight line effect
For example, if I plan my route drawing a line over the map, but I only mark some points, Google will draw a straight line between two consecutive points. Remember that the shortest path between two points is a stright line, any other path will be longer. But when you actually ride the route you hardly ride straight lines, you ride curves. Each curve will add a little more than each line segment in your map, so, when adding all these diferencies up, the result is a longer actual ride than the planned line.
When following or marking a route using some sort of GPS based solution, you have to account also for the "sampling frequency" of the device. For example, lets say you take a minute to travel around a roundabout. If your GPS device also "samples" your position every minute, it will likely mark your position twice in the same spot, so it will record 0 traled distance. This will affect you more in the more curvy trails and routes, specially if you ride at higher speeds. GPS based solutions also work calculating the distance from two consecutive points and adding up with the next pair, and so on. (Yes, it asumes a straight line between points). The higher the sampling fequency the more acurate the distance will be recorded, but also battery consumption will be higher and if files are a concern, they will also be bigger.
This also occurs in the vertical dimension. For example if you side over a small hill first going up a few meters, then going down over an otherwise straight path, but you marked only the beginning and ending point, most likely the software wont take the hill into account, measuring the distance as the straight line between the points.
The curved line effect
There are also another factor that comes into play: We all know that most bike odometers work by counting wheel revolutions. But many of us do not mentally acount for many wheel turns that are not actually accounting for travel distance. Lets say, you lift up the bike and let the wheel spin, or when you get to a resting point and do a few empty loops while waiting for other riders, etc...
Something else is that, in curves, the front wheel actually travels a longer distance than the rear, and you, actually travel a distance that is in between. For example, lets say you ride trhoug a puddle that makes both your tires very wet, an then on a dry surface you ride a really tight U-turn. Now see the path marked by your tires. The fron one must have drawn a bigger semicircle than the rear one. If you want to be sure, measure them with flexible metric tape ;) This seems too little, but over a long route, every tight corner you ride, actually adds up to the distance "seen" from the front tire.
The longer the route, the more this will affect your predicted route distance versus actual ridden distance. But also, the more curvy routes are the ones more severely affected.
To improve the distance acuracy of the planned route, use shorter line segments along curved parts of the trail, i.e. put the points close together.
Note: I have always doubted about wether Google really takes into account ground level when measuring distance over traced paths, I don't know if it includes altitude of the ground at the marked point or if it assumes all points are at the same height. If it asumes same height, calculated line length is less than the actual path.
