Your AI health summary says the risk doubles. From what?

‘The risk doubles’ would make me stop reading. It would not tell me how likely the event is. NICE's existing shared decision-making guidance uses a useful example: a risk increasing from 1 in 1,000 to 2 in 1,000, rather than simply saying it doubles. Those are illustrative numbers from the guidance, not a finding about a particular treatment. NICE also recommends keeping the denominator consistent and giving the relevant time period. If I asked an AI assistant to explain a health article before an appointment, I would want the original absolute numbers beside its summary. Out of how many people? Over how long? In which group? If the article does not supply the starting risk, the assistant should say it is missing—not fill the gap with a plausible number. I would also keep a study's group average separate from a prediction about me. NICE asks professionals to explain how risk information applies to the person and how uncertain it is. A chatbot turning a paper into simpler language has not established my personal risk. This is a reading check I would borrow from clinical communication guidance, not evidence that a particular chatbot passes or fails it. It could leave someone with a better question for their clinician instead of an alarming phrase to carry around all evening. What would you need beside ‘the risk doubles’ before that summary became useful?