In A/B testing, a test hypothesis is a statement you make based on the available data, which serves as the starting point for the test. Before running a split test, you have to set a goal, for example improving the conversion rate. To achieve this goal, you have to take specific steps, so you propose a hypothesis to start with: modifying the text of the landing page, or changing the size of that button will increase sign-ups.
The test hypothesis is only an assumption made in order to test the effectiveness and consequences of a proposed solution. In order to prove or disprove the hypothesis, you need to know what impact you expect from the test, and what changes in metrics you expect to see after implementing the variants.