Abstract
This essay investigates how large language models (LLMs) like ChatGPT and Gemini negotiate legal, political, and cultural scripts associated with migration in the US to construct prototypical immigrant characters and narratives. Given the American security state’s use of commercial LLMs and generative artificial intelligence in immigration management, we conducted a seven-month long empirical study that analyzed synthetic narratives produced in response to four prompt configurations in order to survey attitudes to noncitizens embedded in widely used LLMs. Our research shows that ChatGPT and Gemini, driven by an overarching “ideology of engagement,” simulate the positionality of “social storytellers,” constructing highly patterned stories with widely shared emotional goals to sustain user interaction. Synthetic narratives about immigrants are consistently structured as the “hero’s journey” which privileges happiness, economic mobility, and seamless assimilation while minimizing complex affects, trauma, and structural violence. A benchmark immigrant emerges in our synthetic data in the form of a recurring character: Anya, a young Ukrainian woman who finds joy in the US by opening a bakery. Anya’s whiteness, positivity, and economic assimilation reflect broader cultural myths and biases present in training data and corporate moderation rules. Advancing “Narrative AI Studies,” our essay puts affective narratology in dialogue with natural language processing (NLP) methods to scrutinize story structures, sentiment patterns, and phrase networks. We argue that cheap AI narratives of migration such as those featuring Anya facilitate algorithmic marginalization of real-world migrants. Our research contributes to applied narratology by demonstrating how narrative theory can help assess the political and ethical stakes of AI-mediated storytelling.