Opus 5.5 loves to tell you 'this matters' (and other AI writing tells)
Opus 5.5’s biggest tell is the word “dependable,” which pops up 23 times more often than in human samples.
TL;DR
- Researchers have identified new linguistic tells in AI-generated prose, despite models evolving to mimic human writing.
- A study by Graphite found 13,000 phrases that are significantly more common in AI content than human content.
- Claude models are showing a closer approximation to human word distribution over time, while GPT models are diverging.
- Claude Opus 5.5 frequently uses the word 'dependable' and the phrase 'this matters,' and its variations.
- OpenAI's Astra uses 'another dimension' and 'corrective framing' (e.g., 'not simply X') as prominent tells.
- AI models have significantly reduced their overuse of em-dashes in response to research, but overall tells remain persistent.
- The persistence of these tells suggests that fully eliminating them may be a significant challenge for AI developers.