A tree of questions, ticked over and over
A behavior tree is a way of organizing a robot's decision-making into a tree of nodes, evaluated from the root downward at some fixed frequency, often tens of times per second. Each evaluation is called a tick. When a node is ticked, it does some work and immediately returns one of three statuses: Success, meaning it finished what it was supposed to do, Failure, meaning it couldn't, or Running, meaning it's still in progress and expects to be ticked again next cycle.
That three-way return value is the entire contract of the system. A node never needs to know what its parent is doing with the result, and a parent never needs to know how a child produced its result internally. This strict separation is what makes behavior trees compose well: you can build a small, well-tested node once and reuse it in many different trees without it ever needing to know its surroundings.