Study Finds We Trust 'Expert' Money Advice Over AI — Even When Identical

Scientists wanted to know something simple but important: when a chatbot gives you money advice, what actually makes you trust it? To find out, they ran an experiment with 285 American adults. Each person read eight made-up financial situations — think "should I pay off debt or invest?" — and got a recommendation. The twist: the exact same recommendation was labeled three different ways, as coming from an AI, a human expert, or an online community forum. Sometimes the label was hidden entirely.

The results were striking. The label mattered more than the advice itself. Advice tagged as coming from an expert was rated safer and more knowledgeable, even though the words never changed. The researchers also found that the style of the writing mattered on its own: expert-sounding advice was still the favorite when no labels were shown at all. And these quick gut reactions were powerful — they explained about 69% of how people rated overall quality, 76% of trust, and 83% of whether someone said they would actually act on the advice.

Why should you care? Millions of people now ask ChatGPT, Gemini, or a bank's chatbot for help with budgets, loans, and retirement. This study suggests we're swayed by branding more than substance. A slick "expert-approved" stamp can make mediocre advice feel safe. A small "AI-generated" tag can make good advice feel suspect. That's a recipe for bad decisions — and a gift to scammers who slap authoritative labels on nonsense.

The catch: nobody risked real money here. People reacted to pretend scenarios, so we don't know how they'd behave with their actual savings on the line. The group was also limited to US adults, and trust is partly cultural. Still, the takeaway for AI builders is clear: don't just chase maximum trust. Design tools that help people check the reasoning themselves.