GoKawiilGraphite study catalogs 13,000 AI writing 'tells' across frontier models
Marketing firm Graphite analyzed writing from frontier AI models by having them rewrite 10,000 pre-ChatGPT articles from summaries, then compared the results to the original human text. The study identified over 13,000 words and phrases that appear at least twice as often in AI writing as in human writing, with each model showing distinct habits—Claude Opus 5.5, for example, overuses 'dependable' and phrases like 'this matters.'
GoKawiil's interpretation of the reporting above, not reported fact.
The findings suggest old AI detection heuristics like em-dashes are now outdated as models evolve, meaning detection methods must constantly adapt to new stylistic quirks. Graphite's chief AI officer noted Claude models appear to be converging toward human-like word distributions over time while GPT models are diverging, which could hint at different training or fine-tuning approaches between labs, though the underlying causes remain unconfirmed.
- Graphite found 13,000+ phrases that appear at least twice as often in AI writing versus human writing.
- Each AI model has distinct stylistic quirks, such as Claude Opus 5.5's overuse of 'dependable' and 'this matters.'
- Claude models are reportedly trending closer to human word patterns over time, while GPT models are trending further away, per Graphite's chief AI officer.
Source: techcrunch.com — Russell Brandom, 2026-10-01
Published there as: “Opus 5.5 loves to tell you ‘this matters’ (and other AI writing tells)”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.