Register mismatch

Every language layers its vocabulary by situation — academic, professional, casual, intimate. Those layers do not line up across languages. A word that is neutral in your first language can map to an English "equivalent" that is markedly casual or oddly stiff, and dictionaries rarely warn you which is which.

ESL writers hit this from both directions. Downward: conversational phrasing lands in an academic paragraph — "The results were a big deal for the field" → "The results were significant for the field." Upward, the more common failure: vocabulary lists push learners toward heavy formal words that native writers rarely stack together — utilize, moreover, plethora, thus, all in one paragraph. Both directions are grammatical; both read as off.

Register is about fit, not correctness. A grammar checker will pass a mismatched sentence because nothing in it is technically wrong. That is why register errors survive proofreading tools that reliably catch agreement and article mistakes — and why they need a human ear, or a tool that understands context, to fix.

This matters because readers judge register before they consciously judge anything else. A reviewer may not articulate why a paper feels unpolished; often the answer is three casual phrases in an otherwise formal text.

There is an AI-detection angle too. The over-formal, wordlist-driven style that ESL instruction rewards overlaps heavily with the default style of large language models. A human writer who learned English from academic word lists can sound "AI-like" to both human readers and detectors — one more reason to treat automated flags on non-native writing with skepticism.