Field notes
How AI writing gives itself away
The tells we catalogue, why they show up, and what to do about them. No listicles about the future of work.
26 August 2026 · 10 min read
AI detectors flag human writing constantly, and here is why
Detectors do not detect authorship. They measure how predictable your word choices are and how much your sentence lengths vary. Plain, well-structured, professionally edited writing scores as predictable, which is why it flags. Seven detectors ran over 91 TOEFL essays written by non-native...
ai detectionfalse positivesclient work
26 August 2026 · 12 min read
The 908 words AI overuses, and what to write instead
"Delves" appears in biomedical abstracts at 25.2 times its expected rate, measured across 15 million papers by Kobak and colleagues in Science Advances in 2025. It is one of 908 terms our catalogue tracks across 18 categories. No single word convicts a draft. Six of them in one paragraph does....
ai wordsvocabularyslop lexicon