Product changes to earn user confidence in Bing's medical search answers

You are a product manager on Microsoft's Bing team. People come to Bing with health questions about symptoms, conditions, medicines, treatments and where to get care, and a misleading result on these topics can do real harm. How would you increase how much users trust the results Bing shows for healthcare searches? Explain which problems you would go after and what you would build first. ```hint Trust has two sides Separate whether the results deserve trust from whether users feel they can trust them. A change that moves only one of the two can backfire. ``` ### Clarifying Questions - Is there evidence of a trust problem today, such as low survey scores, health queries that users keep reformulating, or complaints about misinformation, or is this a proactive investment? - Which surfaces are in scope: the ranked web results, sponsored results, answer panels, or the AI chat answers shown alongside search? - Which markets and which kinds of health queries matter most: symptom checks, medication questions, chronic-condition research, or finding a doctor? - Is success judged by users' perceived trust, by the clinical accuracy of what is shown, or by Bing's share of health searches? ### What a Strong Answer Covers - A definition of trust for health search that separates how reliable the results actually are from how credible users find them. - A chosen user segment and the specific moments where trust breaks, such as contradictory sources, ads that look like results, alarming worst-case pages, or AI answers with no visible sources. - Concrete changes across ranking, source transparency, AI-answer safeguards and urgent-symptom handling, prioritized with a stated reason. - Metrics that combine expert-rated accuracy with behavioral and survey signals, plus guardrails on query volume, revenue and fairness to publishers. - Risks such as liability, suppressing legitimate sources, and keeping up when medical guidance changes. ### Follow-up Questions - How would you measure trust in a way that cannot be inflated simply by adding reassuring labels? - During a fast-moving outbreak, official guidance changes week to week. How would ranking and AI answers keep up without amplifying rumors? - Health-related ads are a revenue source. How would you weigh stricter ad rules against that revenue?