Narrative construction of "AI" through anthropomorphising language in German public discourse

This paper examines narratives about “AI” as a human-like actor in the German public discourse across media and over time. We define narratives as recurring lexico-grammatical patterns that anthropomorphise AI and analyse 41,358 deduplicated sentences containing the word AI from quality newspapers, IT blogs, and web texts (2019–2024). Combining spaCy-based NLP with manual annotation, we investigate: (1) anthropomorphic lexemes in the immediate context of AI (agent, cognizer, communicator, biological metaphor), (2) the syntactic role of AI (subject vs. object), and (3) pronominal reference to AI. Results show that agency is the dominant category, with newspapers becoming the primary locus of anthropomorphising language after the release of ChatGPT. Pronominal reference to AI remains rare but increases slightly over time, especially in the press. We argue that these patterns constitute “distantly read” AI narratives that shape expectations about AI’s autonomy, cognitive, and communicative capabilities.