Researchers have identified thousands of new linguistic clues that can indicate whether a document was written by artificial intelligence, according to reporting from inc.com. The findings, covered across several technology outlets, point to specific habits in AI-generated text that readers and editors can learn to spot.
Among the most discussed tells is a fondness for the phrase "this matters," which TechCrunch reports is a favorite of the model Opus 5.5. VentureBeat reports that Claude Opus 5.5 uses em-dashes 99 percent less often than earlier versions and reads as more human, yet still displays 2,548 distinct AI writing tells. Gizmodo describes the work as identifying thousands of new tells, quoting researchers who say the subject "matters."
The coverage suggests that detection is shifting away from a single giveaway, such as a particular punctuation mark, toward patterns of phrasing and emphasis that recur across a model's output. As one habit is reduced, others remain, which is why researchers continue to catalog them in large numbers.
The reports come from inc.com, Gizmodo, TechCrunch and VentureBeat, all published on October 5, 2026. The same news summary also included an unrelated item from a Ukrainian outlet about Kyiv's mayor reporting a building strike and a separate warehouse fire in the Obolonskyi district.
For anyone reviewing a suspicious document, the practical takeaway is that no single clue settles the question. What to watch next is whether these catalogs of tells keep pace as models are updated to sound more human.
