Even with careful attention to these issues, it is necessary to recognize the fundamental limitations of generative AI. Traditional search engines—despite the problems inherent in their search algorithms—present users with lists of sources from multiple websites. By contrast, generative AI typically provides a single, consolidated answer, which increases the risk that users may accept AI-generated responses uncritically. Moreover, bias does not arise only in relation to expressions concerning socially marginalized groups. Generative AI may also exclude non-mainstream perspectives within a society, as well as viewpoints or information that are not well represented or expressed on the internet. When these structural issues are taken into account, merely subjecting generative AI outputs to critical review may not be sufficient. For this reason, it is essential to avoid overreliance on generative AI. Particularly when dealing with important topics, organizations should always keep in mind the need to gather information and perspectives through diverse channels, such as direct research and consultation with experts. 72 73