I'm Romanian. I've spoken English for twenty years, I write it every day at work, and I dream in it about half the time now. But when I'm under pressure — applications, cover letters, the high-stakes writing where I most want to sound polished — I revert to patterns from Romanian that are grammatically correct and subtly non-native. My sentences are right. They just sound like they're right, in a way that's hard to name and easy to feel, and that gap was quietly costing me interviews.
What a native speaker hears that I genuinely cannot
The maddening thing is that the tells are invisible from the inside. I proofread my CV a dozen times and found nothing, because to me it read fine — of course it did, I wrote it with the same brain that produced the tells. I needed an outside ear. So I put my CV summary into the Make My Non-Native Writing Sound Like a Native Wrote It prompt, told it my native language was Romanian and the register was a professional cover letter, and waited to feel exposed.
The TOP 3 PATTERNS section was specific and immediately, embarrassingly recognizable. The first one: Romanian speakers often drop or misplace definite articles before abstract nouns in English, because Romanian handles that grammatical job differently. My 'Experienced in project management and team leadership' wasn't wrong, exactly, but a native would more naturally write 'I've managed projects and led teams across…' — less nominalized, more verb-driven, more alive. Once I saw it I couldn't unsee it; I'd been writing in nouns where English wants verbs my whole career.
I used Grok 4.1 Fast Reasoning here on purpose. This is a fast, iterative, do-it-again-on-the-next-paragraph kind of task — I ran my whole CV through it section by section — and the low-latency reasoning meant I could blast through it in one sitting without waiting around. It's the right tool when you want quick, sharp edits in a loop rather than one big careful pass.
The rewrite was LESS formal, which was the real lesson
Here's what I didn't expect: the native version was less formal than my original, not more. I'd done the classic non-native overcorrection — assuming that more elaborate, more formal, more Latinate words sound more competent. They don't. They sound like someone trying. The prompt explicitly refuses to over-formalize, and the rewrite stripped my stiff, defensive formality back down to clean confident plain English. I'd been armoring my writing against judgment and the armor was the thing getting judged.
- Article usage and noun-heavy phrasing are the most common tells for many European-language speakers writing English
- The TOP 3 PATTERNS list is the real value — read it and you start catching the pattern yourself next time
- Non-native writers tend to over-formalize; a good edit often makes the text simpler, not fancier
- Feed it the high-stakes writing (CV, cover letter) where your tells spike under pressure
- Setting your native language tells it which specific tells to hunt — Romanian transfer errors differ from Mandarin ones
Why telling it my native language matters so much
The single input that makes this prompt smarter than a generic grammar checker is the writer's-native-language field. Non-native tells aren't random; they're systematic transfers from your first language, and they're completely different depending on what that language is. A Romanian speaker drops articles in one specific way; a Mandarin speaker drops them in a different way and also tends to flatten verb tenses; a Spanish speaker over-uses certain prepositions because they map one-to-one in Spanish and not in English. When I told the prompt 'Romanian,' it stopped looking for everyone's mistakes and started looking for mine, and the TOP 3 PATTERNS were uncannily on target. A generic checker would have caught the same errors as 'wrong' but never explained them as a coherent accent, which is the difference between fixing one sentence and fixing yourself.
I leaned on Grok 4.1 Fast Reasoning because I wasn't doing one careful pass, I was doing twelve fast ones — running every section of the CV, then my cover letter, then a couple of LinkedIn paragraphs, in a tight loop, reading the pattern notes each time and watching them start to repeat. By the third or fourth pass I was predicting the note before it appeared, which is exactly when you know the lesson has transferred from the tool into your own head. The low latency mattered: a slow model would have broken the loop and the loop is where the learning happened.
The suspicious results
I sent the revised CV to three roles the following week and got two callback interviews, which beat the previous two months combined. I cannot prove causation — maybe the market shifted, maybe I got lucky, maybe the moon was right. But I have a strong suspicion, and the suspicion is shaped exactly like a missing definite article. Now I run important writing through this before it goes out, and when I'm specifically composing in another language from scratch rather than fixing English, I start with my Write a Native-Sounding Business Email in Any Target Language prompt instead.
If your second-language writing is correct but somehow not landing, grab the Make My Non-Native Writing Sound Like a Native Wrote It prompt on Prompt Dock, paste in the thing you're most nervous about, and read the top 3 patterns. Those three patterns are probably most of your accent on the page.