When it comes to health issues, trust in AI chatbots is just 7 percentage points behind search engines. And the most trustworthy sources are the least frequently used.
A severe sore throat and a slight fever for the past two days. Should I see a doctor, or will home remedies suffice? We presented this scenario to around 300 working people in German- and French-speaking Switzerland as part of our study on AI adoption and asked: How trustworthy are the various online sources in this situation?
The result
Respondents trust the websites of pharmacies, hospitals and doctors (92 per cent) as well as health portals (90 per cent) the most. These are followed by search engines such as Google at 82 per cent and, just behind them, AI chatbots such as ChatGPT at 75 per cent. Trailing far behind are YouTube (37 per cent), forums (35 per cent) and social media (21 per cent).
So only 7 percentage points separate the AI from the search engine. When it comes to strong trust (‘very’ or ‘extremely trustworthy’), the gap is just 4 points: 33 per cent for Google, 29 per cent for the AI.
It’s not that AI is winning; it’s that Google is losing
Things get interesting when we compare this with our second scenario: a simple DIY question. In that case, 91 per cent trust the search engine, whereas for the health-related question, the figure drops to just 82 per cent. YouTube’s trust rating even plummets from 86 per cent to 37 per cent. Trust in AI chatbots, on the other hand, remains almost unchanged: 71 per cent for DIY, 75 per cent for health.
AI is perceived as a generalist. Not as credible as a specialist source, but not rejected either – regardless of the topic.
Convenience trumps trust
Perhaps the most important finding: the most trustworthy sources are the least likely to be used first. Only 19 per cent cite the websites of pharmacies, hospitals or doctors as their first or second port of call. For AI chatbots, the figure is 60 per cent.
People start where they can get an answer most quickly. Only afterwards do they check specialist sources, if at all.
What this means
For insurers, healthcare providers and public authorities, it is no longer enough simply to have the best information. It must reach people where they look first: in search results and, increasingly, in AI responses. And if three out of four working people at least tend to trust AI when it comes to health questions, the quality of these answers becomes a matter of public health.
In the full study report, you will find all the results broken down by age, language region and scenario, along with the methodology and tables.
→ Download the study report: https://www.data-iq.ch/download-studienbericht-kinutzung-doi