An A/B Test allows you to randomly display different variables (sometimes to referred to as stimuli). These variables can be versions of an image (generally a logo or product design) or text (such as a slogan). You set how often each version will be shown to respondents. Then when you analyze your results, you can assess how the different stimuli affected responses.
As you add tests, the percentages of how each test will be shown is split evenly across all tests. You can also enter your own percentages. The total of all percentages must sum to 100%. These percentages determine the probability of a new respondent seeing the corresponding test when they open up your survey.
Most often two variables are used, split 50%/50%. For example you could show two versions of a new company logo. The question that follows the logo would ask "How does the logo make you feel?" You would be looking for one logo to elicit a more positive response than the other. 50% would see logo A, and 50% would see logo B.
Each time you preview the survey, a random A/B test will be shown based on the percentages you specified. You should preview your survey a few times to make sure you know how each A/B test variables will look to respondents.
You could also create a comparison report for the A/B test. This would seperate each question by A/B test. So if you asked "How does this logo make you feel?", you would see that question broken out into respondents who saw Logo A, and respondents who saw Logo B.