This study examines how Large Language Models (LLMs) respond to migration-related survey instruments. Survey items will be selected from the European Social Survey Round 12 (2025/26) rotating module on ‘Attitudes to immigrants and refugees’, supplemented by migration-related items from the ESS core questionnaire. These items will be administered to a small sample of LLMs (such as ChatGPT, Claude, and DeepSeek), selected to capture variation in institutional origin, training context, and openness. Responses will be mapped to the original survey response scales, enabling two levels of comparison: cross-model analysis to identify systematic differences between LLMs, and benchmarking against ESS population-level data to examine how AI-generated responses align with or diverge from European public opinion.

Duration: 2026 – 2027

Funding: self-funded

Principal Investigator: Angelo Tramountanis

Research Team: Till Hanauer