Science journalism has developed a house style for AI stories: AI helps, AI improves, AI suggests. The frame is structurally affirmative — each story positions AI as a tool that expands what researchers can do — and it has held even as the researchers closest to the work have grown measurably cooler on its conclusions. The mismatch is not a matter of journalists getting facts wrong. It is a matter of emphasis: which findings make the lede, which caveats get buried, and whose voice gets quoted as the authoritative close.…