A/B testing (aka split testing) allows you to compare multiple versions of a page to see which one converts better. To be a true A/B test, you must test only one element at a time (e.g. long copy vs. short copy, images vs. no images, red button vs. blue button, etc.) For your testing results to be reliable you must run the test until you reach statistical significance.
"Being able to test assumptions in a meaningful way is what A/B testing is about."