A mile high and an inch deep: Exploring <scp>ChatGPT</scp> as a mathematics curriculum development tool

Abstract Artificial Intelligence (AI) technology has issues, including inherent bias and inaccurate information, yet it has not stifled individuals' use of this technology in the mathematics classroom. Thus, we conducted a document analysis to investigate ChatGPT, a generative AI chatbot, constructing elementary mathematical tasks to determine how mathematics teacher educators (MTEs) can support this implementation. We analyzed 191 text responses constructed by ChatGPT, each corresponding to a task created for one of the elementary Common Core mathematical standards from kindergarten to fifth Grade to answer the research questions: What are the associated levels of cognitive demand of the ChatGPT's text responses, is there a statistically significant difference between lower‐level mathematical tasks and higher‐level mathematical tasks, and what are the common characteristics of AI's tasks? We found significantly higher‐level cognitive demands across the tasks, but many activities were repetitive and inappropriate for the specified grade level. Finally, we explored how MTEs could support teachers using this tool in their classrooms using Mathematical Critical Curation Questions.