Opus 5.5 Loves to Tell You 'This Matters' — And Its Biggest Tell Is 'Dependable'

TechCrunch reports that Opus 5.5 has a set of distinctive writing tics, most visibly a habit of telling readers that “this matters,” while its single sharpest marker is the word “dependable,” which shows up roughly 23 times more often than in human samples.

The idea of an “AI writing tell” covers the features that make a passage read as machine-written, whether that is a recurring word choice or a repeated sentence shape and rhythm. Individually these signals are easy to miss; together they form a recognizable fingerprint.

Those fingerprints matter because they bear directly on how identifiable AI text is. Content provenance, academic integrity, and a reader's judgment about who actually wrote something all rest on the ability to separate human writing from machine output.

Tells emerge from how models are trained. Large corpora and alignment processes push models toward shared preferences in phrasing, and the more a particular construction is reinforced, the more often it resurfaces, hardening into a predictable habit.

The result looks like an ongoing tug of war. Model makers want output that reads more naturally and less mechanically, while researchers and detection efforts keep hunting for fresh signals. When one side tightens, the other shifts to a new feature.

What to watch next is whether tells like these fade in the next generation of models or simply move to new words. For writers and platforms, the more durable approach is to track shifts in overall text patterns rather than fixate on any single giveaway keyword.