generated solely from the model’s internal training data. However, because links cited by generative AI may be broken, outdated, or based on sources of limited relevance or credibility, it is necessary to verify the accuracy of sources one by one even when the output claims to be based on external information. It is advisable to prioritize official documents, authoritative sources, and academic research relevant to the topic. At the same time, it should be recognized that reports published by governments, international organizations, or public institutions may also reflect politically biased perspectives or include distorted data. Because laws may be amended and specific events may evolve over time, it is also necessary to check whether more up-to-date information is available. Given that such verification work requires significant time and effort, there may be cases in which using generative AI is, in fact, less effective rather than more. Another possible approach is to pose similar questions to different generative AI systems and compare their responses. Because these systems may rely on different sources, any discrepancies in factual details should be treated with particular caution. Ultimately, the responsibility for making a final judgment about AI-generated outputs lies with the organization and the activists responsible for the work. Making sound judgments requires the experience and expertise of those individuals. This is precisely why activists’ capacities remain essential even when generative AI is used. If those responsible lack sufficient expertise, even supplementing AI outputs with internet searches or expert Generative AI Guide for Civil Society