Where the term shows up In this channel, AI tell appears during live model reviews — notably the GPT-6.1 Soul vibe check — as the label for writing that immediately gives a model away. The reviewer flags flat, unnatural, near-equal sections as an easy AI tell, even while praising the same model's benchmarks, computer use, and visual output. It functions as the counterweight in a vibe check: a model can score well and still fail the human read. How it is used in practice During the DevDay 2026 review, the verdict splits cleanly: excellent design and visual tasks versus writing that is not a reason to pick this model. The AI tell is what the reviewer points to when summing up that weakness — segmented prose that reads as machine-produced at a glance. - AI tell = visible machine-flavored writing, here described as flat, unnatural, near-equal sections.- It surfaced specifically in GPT-6.1 Soul's writing, alongside otherwise top benchmark scores.- It's a practical signal in a vibe check: how a model writes can disqualify it for writing tasks regardless of benchmarks.