B-roll is the footage that supports your main video: the product close-up, the screen recording, the diagram that appears while you keep talking. AI changes how that supporting material gets found and placed, and it changes it in four different ways that should not be treated as identical. AI can suggest where a visual would help, select existing footage from a library, adapt material to fit a frame, or generate entirely new images that never happened.
That last difference matters more than any tool choice. A generated clip can illustrate an abstract idea beautifully and it can fabricate evidence terribly. The rule that runs through this whole series: B-roll should support the message rather than merely create movement, and a visual must never make a stronger factual claim than the words it covers.
The purpose of B-roll is not to make the screen constantly change. Sometimes the strongest available visual is the speaker's own face, and the right amount of B-roll is none. These five guides cover the craft end to end: what B-roll is, how to add it, when AI beats stock (and when neither fits), how to keep a talking head human, and 36 concrete examples sorted by what you are trying to say.