Hallucinated references and Indigenous voices

A common issue with the use of AI for research is the hallucination of references, sometimes called hallucinations. These are fake or fabricated citations that may appear to be very real. This occurs when the generative artificial intelligence (Gen AI) tool attempts to give you what you want by identifying patterns in information that seem likely to be true. They may be the names of real scholars, or they may be real last names and incorrect first names; they will be perfectly formatted with titles and journal names (sometimes real) and a fake doi (digital object identifier). Checking the doi is one of the easiest ways to identify an AI hallucination. Currently, approximately 40% of citations offered by an AI chatbot are hallucinated (Jones, 2025). The percent of hallucinations can be better or worse depending on the application you are using. Occasionally, the article will exist, but its content will be completely different from what the GenAI chatbot stated.

Hallucinated references are an issue under any circumstances; however, when applied to Indigenous Knowledges, they can mimic the harms done by real life person’s committing identity fraud (for details on identity fraud, see chapter 3). Chief among these harms are misinformation, the misrepresentation or appropriation of culture, the displacement of real Indigenous scholars, and the erasure of context and diversity.

Already in the habit of being careful when selecting and evaluating resources, Indigenous students and scholars may find that they do not have to greatly alter their behaviours. Under regular circumstances, an Indigenous scholar will follow the process of attempting to verify the identity of the authors they are hoping to cite. Being equally as thorough with AI references is important. Alterator, S., Maddox, A., Southerton, C., & Schutt, S. (2025) found in their study Re-authorising Colonialism or re-authorising Indigenous perspectives: The Risks of Generative AI in Indigenous Education contexts that to critically evaluate information a researcher would need an understanding of the historical and current realities of Indigenous Peoples, an awareness of deficit discourses and an awareness of their own positionality and this is challenging for many researchers and currently an impossible task for GenAI.