This paper investigated how users trust generative AI and how this influences their decision-making. In human-AI collaboration, over-reliance or under-reliance on AI can lead to lower decision-making quality than without AI support. On the other hand, with generative AI, users can ask various questions and receive responses derived from extensive training data, enabling more dynamic interaction. In this context, it remains unclear how users form trust in generative AI and how this trust influences their decision-making. To explore this, we conducted a human-generative AI collaborative decision-making experiment using a semi-structured problem that cannot be completely solved by mathematical formulation alone. We quantitatively and qualitatively analyzed multiple aspects of collaborative decision-making outcomes, processes, and users’ trust in generative AI. In the results, we found that the use of generative AI tends to cause confirmation bias, which could lead to a reduction in the quality of collaborative decision-making. However, by collaborating with users from the initial stage of the decision-making process, which is problem recognition, confirmation bias may be suppressed. This paper provides some fundamental findings that could enhance effective human-generative AI collaborative decision-making, particularly in the domain of trust.