Machine-Generated Text

Machine-generated text is the broader category that includes AI generated content but also covers older automated systems: template-based article spinners, database-driven product description generators, and rule-based text production systems. The term predates the modern LLM era but has taken on new urgency as large language models made machine-generated text indistinguishable from human writing to casual readers.

For detection purposes, machine-generated text from modern LLMs is qualitatively different from older automated content. Spinners and template systems produce text with obvious mechanical signatures — unnatural phrasing, poor coherence, repetitive patterns. LLM output is fluent and coherent, which is precisely what makes detection difficult and why statistical approaches (perplexity, burstiness) became the primary detection method.

The regulatory and policy environment around machine-generated text is evolving. The EU AI Act includes disclosure requirements for AI-generated content in certain contexts. Academic journals are developing explicit author statements about AI use. Content platforms are setting their own policies. For producers of AI-assisted content, understanding which disclosure requirements apply to their specific context — and meeting them honestly — is increasingly a legal and reputational question, not just a technical one.