Show HN: Training a model to identify AI web content from structure alone

Category: ai-ml

Tags: ai-detection, structural-features, replication-study, nlp, research-package

Score: 6.3/10 (Innovation: 7, Technical: 6, Documentation: 7, Utility: 5)

SlopShape is a research release that replicates the StoryScope structural-detection study on commercial B2B blog content, showing that AI-generated posts are identifiable from 203 structural features alone at 97.0 macro-F1, nearly unchanged after rewording. It provides a fully reproducible pipeline, verification artifacts, and a frozen instrument, but gates the per-document scoring data and is noncommercially licensed.

Target audience: AI researchers, content authenticity researchers, and platform trust-and-safety engineers

Repository: https://arxiv.org/abs/2609.15369 · Python · NOASSERTION · 74 stars

View on Hacker News