From Computer Science Wiki
Advanced programming[1]

Elitism is a method where the most fit individuals of a generation, usually a small number compared to the population, are guaranteed a place in the next generation without alteration in their parameters or values in order to avoid decline in quality that may result from crossovers or mutations, although sometimes with a second copy of themselves with mutations will be present as well, and in some cases entirely bypassing the mating process that less fit but still fit enough individuals after selection have to go through with the goal of generating a succeeding generation that ensures an improvement in quality or behaviour, or whatever is evaluated for 'fitness'.

Example: In the travelling salesman problem, the top three routes in a generation may move on to the next generation, making sure there will be some good individuals in case crossovers and mutations really mess up the pathing.

Example: In an evolution simulator, the best individuals may wait until the next generation to reproduce so their genes don't get scrambled.