Graphite finds 13,000 phrases distinguishing AI writing from human
Graphite, a marketing firm, analyzed AI-generated prose to identify persistent linguistic patterns distinguishing it from human writing. The study compared 10,000 pre-ChatGPT articles with AI rewrites of those summaries to minimize bias. Researchers found 13,000 phrases at least twice as common in AI content, including model-specific quirks like Claude Opus 5.5’s overuse of “dependable” (23x…
Key points
- Graphite identified 13,000 phrases at least twice as common in AI-generated writing vs. human samples
- Claude Opus 5.5 overuses ‘dependable’ (23x) and ‘this matters’ (116x) in its output
- OpenAI’s Astra favors ‘another dimension’ phrasing and hedges claims with ‘may provide’
Model pages: Claude Opus 5.5 → · Astra →
Opus 5.5 loves to tell you ‘this matters’ (and other AI writing tells)
TechCrunch AI · 1 October 2026
Loading the full article…
This text was published by TechCrunch AI and written by Russell Brandom. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Generative AI & Models
All →- Google releases Gemini 4 Argon with 1M token output limit · 19 src
- Tavus' Griffin AI avatar fools 48% of users in one-minute call · 1 src
- Ideogram introduces Ideogram 4.5 for precise image editing · 1 src
- OpenAI releases GPT-6.1 Sol, says it nearly matches Astra on agentic coding at one-fifth the price · 35 src
- Anthropic releases Claude Sonnet 5.5, faster and more cost‑efficient model · 4 src
Comments
via GitHub Discussions