Attribution Gradients: Incrementally Unfolding Citations for Critical Examination of Attributed AI Answers