The Formulaic Machine: Generative AI, Academic Writing, and the Recovery of Voice

This paper examines the growing influence of generative AI on academic writing, with particular attention to the formulaic phrasal patterns AI tools introduce into scholarly prose. Drawing on observations made at an international humanities workshop (MEMaT, Rome), the author—a computational linguist with a background in the language modeling research that preceded today's large language models—identifies a set of recurring stylistic signatures characteristic of AI-generated academic text. The paper argues that while generative AI offers genuine benefits—particularly for non-Anglophone researchers—heavy reliance on these tools risks flattening distinctive scholarly voices and eroding the intellectual labor that academic writing is meant to embody. Practical recommendations are offered for writers who wish to recover their own voice. The paper is itself an experiment in the form it analyzes: the main body was drafted by Claude (Anthropic), with the human author supplying the prompt, the framing, the introduction, and concluding remarks—demonstrating both the capabilities and the characteristic limitations of AI-generated academic prose.