Opus 5.5 loves to tell you ‘this matters’ (and other AI writing tells)

A new study by the marketing firm Graphite examined the writing habits of frontier AI models, identifying characteristic words and phrases that distinguish their prose from human writing.

Graphite started with a corpus of 10,000 articles published before the release of ChatGPT as a human-generated control group, then had different AI models rewrite the articles from summaries to reduce source bias. It found 13,000 phrases that were at least twice as common in AI content as human content, which it calls a "tell."

According to the results, Claude Opus 5.5's biggest tell is the word "dependable," which appears 23 times more often than in human samples. The model tends to use the construction "is more than an X, it's a Y." It uses "this matters" 116 times more often than human writing, and "why X matters" 92 times more often.

OpenAI's Astra has different tells, including describing "another dimension" of a topic, hedging claims with "may provide" or "can provide," and what Graphite calls "corrective framing," defining a topic as "not simply X" or offering an alternative "rather than relying on X." Those constructions were more than 100 times more common in Astra-generated prose than in human writing.

All the frontier labs appear to have reduced em-dash use. In Graphite's samples, Opus 5.5 used the punctuation 99% less often than Opus 5. Astra uses it 88% less than human samples, while Gemini 3.1 Pro has almost completely eliminated it.

Graphite's chief AI officer Greg Druck told TechCrunch that Claude models are getting closer to the human word distribution over time, while GPT models are getting further away. He also said the overall number of tells is mostly holding steady, and that labs may be less able to control them than expected because the models have billions of parameters and a finite number of tests.

Anthropic said in the Opus 5.5 release that the model "communicates more naturally than prior models," and OpenAI said users of the GPT-6 versions of Sol and Luna could "expect to see more clarity, less jargon, [and] fewer odd turns of phrase."