Trust in science has dropped below its pre-pandemic level, Wellcome's global survey finds

When a doctor or nurse recommends a vaccine, the most common response in Wellcome's survey is to wait. Asked what they do, 48% said they would prefer to think about it, 34% said they wanted it immediately and 15% said they do not get vaccines at all ^[4]. In that question the person making the recommendation is a clinician, in person. We think a "clinically proven" badge on a signup screen asks for the same act of trust from much further away.

Most teams, in our view, treat a "backed by science" line as credibility that costs nothing to add. The survey measures how many people would grant it. Strong trust is a minority view: 26% of people globally said they trust science a lot ^[9]. The broader measure, at least some trust, is five points under its 2018 level ^[16].

Income is where the pre-pandemic baseline matters most. Strong trust fell further among people struggling financially than among the comfortable ^[7], so the gap between the two groups widened by six percentage points from 2020 to 2025 ^[17]. John-Arne Rottingen, chief executive of Wellcome, said people short of money may see science as part of a broader system that is not working for them ^[10]. "Science is seen as one of those institutions that are not delivering," he said ^[11]. The survey does not establish what caused the overall decline. Rottingen said the politicisation of science during the pandemic probably contributed ^[13].

The hospital figure is the one we would put in front of a product team. Among people with a lot of confidence in their country's hospitals and clinics, strong trust in science reaches 43%, against 26% overall ^[9]. The link is an association, and the survey does not show that good clinic visits create trust. We'd still expect the same pattern with products: a claim a user has seen pay off in their own account carries more weight than one they are asked to accept on authority. "We haven't been good enough at explaining and engaging with those groups and those communities on how science is benefiting them," Rottingen said ^[12].

US teams get a different picture. The headline number rose there ^[14], but among Americans who disapproved of their country's leader, 74% said science benefited people like them "a lot", against 44% among those who approved ^[15]. The fieldwork ran after Donald Trump won the 2024 election, while Joe Biden was still president ^[5].

The test we'd apply has two axes. One asks whether a product's central claim rests on the authority of science or on a result the user can check inside the product within their first weeks. The other asks whether the people buying are mostly comfortable or mostly finding it hard to get by. Authority claims sold to comfortable users draw on the 39% of that group with strong trust. Sold to stretched users, they draw on 20% ^[6]. We'd expect checkable claims to hold up better in both rows. For teams selling to stretched users, our recommendation is to move the citations down the page and lead with what the product shows in use. The tradeoff is time. A result in the user's own account arrives later than a badge does, so time-to-value has to be short enough to carry the first week.

One forcing function makes the choice concrete. Strip the word "science" and every study citation from the onboarding copy and read what is left. If the remaining screen still gives a reason to stay, the research is supporting evidence. If it does not, the pitch rests on the 26% of people who trust science a lot ^[9].