The Language You Ask In: Language-Conditioned Ideological Divergence in LLM Analysis of Contested Political Documents

TL;DR

The study demonstrates that prompting an identical LLM with Russian versus Ukrainian can yield substantially different ideological orientations when analyzing the same political content, exemplified by a 2019 Ukrainian civil society joint statement. Using a controlled, parallel-prompt design with ChatGPT 5.2 on the same document, the Russian prompts elicit a discourse aligned with state-centric, anti-democratic framings, while Ukrainian prompts echo Western liberal-democratic vocabulary and legitimizing civil society oversight. The findings illuminate mechanisms of cross-lingual bias transfer, potential linguistic relativity effects, and the risk that AI outputs could reinforce information warfare framings in multilingual contexts. They further argue for multilingual auditing, transparent reporting of prompt language, and methodological refinements to ensure responsible AI deployment in politically sensitive settings.

Abstract

Large language models (LLMs) are increasingly deployed as analytical tools across multilingual contexts, yet their outputs may carry systematic biases conditioned by the language of the prompt. This study presents an experimental comparison of LLM-generated political analyses of a Ukrainian civil society document, using semantically equivalent prompts in Russian and Ukrainian. Despite identical source material and parallel query structures, the resulting analyses varied substantially in rhetorical positioning, ideological orientation, and interpretive conclusions. The Russian-language output echoed narratives common in Russian state discourse, characterizing civil society actors as illegitimate elites undermining democratic mandates. The Ukrainian-language output adopted vocabulary characteristic of Western liberal-democratic political science, treating the same actors as legitimate stakeholders within democratic contestation. These findings demonstrate that prompt language alone can produce systematically different ideological orientations from identical models analyzing identical content, with significant implications for AI deployment in polarized information environments, cross-lingual research applications, and the governance of AI systems in multilingual societies.