The real problem isn’t grammar

Here’s the thing nobody tells you: your grammar is probably fine. Not perfect — nobody’s is — but fine. The mistakes that actually cost non-native writers readers are almost never missing articles or a wrong preposition. They’re tone mistakes.

You write “I would like to kindly request” in an email where a native speaker would write “Can you send me…?” You use a formal word where the situation wants a casual one, or a casual one where the situation wants formality. You pick the technically correct word and the slightly wrong feeling. Readers don’t think “bad grammar.” They think “this feels stiff” or “this feels off,” and they stop reading.

So the first rule of using AI tools as a non-native speaker: stop asking the tool to fix your grammar. Start asking it to check your register. “Does this sound formal or casual?” “Would a native speaker say this to a colleague, or to a professor?” That’s the question that actually moves your writing forward.

Use the tool as a translator of intent, not a rewriter

The most useful way I know to use an AI tool in a second language is backwards from how most people use it. Don’t write in English, then ask the tool to fix it. Write what you mean — in your own language if that’s easier — then ask the tool to render it in natural English.

I tested this with a friend who writes product copy in English but thinks in Urdu. Her English drafts were correct and lifeless. When she started writing the idea in Urdu first — the joke, the emphasis, the exact point — and asking the tool to carry the meaning across, the English got better overnight. Not because the tool is a great translator. Because her thinking was clearer in her own language, and clear thinking survives translation. Muddled thinking doesn’t.

Try it once: draft the tricky paragraph in the language you think in, then ask for a natural English version, and compare it with the English you would have written directly. Keep whichever one sounds more like you. Most of the time, it’ll be the translated one.

Grammar checkers: trust them, then argue with them

Grammar tools are genuinely useful for non-native writers — more useful than for native ones, honestly. They catch the article you dropped, the verb that doesn’t agree, the preposition that no native speaker would choose. Use one. Run every important draft through it.

But here’s the catch: grammar tools are conservative. They prefer the most common sentence, the safest construction, the phrase that appears a million times in their training data. If you accept every suggestion, your writing gets more correct and more boring in exact proportion. And “boring” is a bigger problem than a missing “the.”

My rule: accept the fix when you understand the rule behind it. If the tool changes “informations” to “information,” take it — that’s a real rule, and now you know it. If the tool wants to replace your perfectly clear sentence with a longer, flatter one, reject it. Your sentence was fine. The tool just prefers beige.

And when a suggestion confuses you, don’t just click accept. Ask the tool why. “Why is this wrong?” A two-sentence explanation teaches you more than a hundred silent corrections, and next month you won’t need the tool for that mistake at all.

The over-correction trap

This is the one that worries me most, because it happens slowly and it feels like progress. You run every sentence through the tool. The tool smooths everything. Your writing gets cleaner and cleaner — and less and less yours. After a few months, you can’t tell whether a sentence sounds like you or like the tool’s default setting.

I see this in edited drafts all the time: the writer’s original had a slightly unusual word choice that was actually vivid and correct. The tool replaced it with the ordinary word. The writer accepted it because the ordinary word felt safer. Multiply that by a thousand sentences and you’ve erased the most interesting thing about your English — the fact that you see the language from outside it.

Non-native speakers notice things native speakers never do. You hear the music of a phrase instead of just its meaning. That outsider’s ear is an advantage, not a defect. Don’t let a tool sand it off.

The practical fix is simple: keep one sentence per paragraph that the tool didn’t touch. Not a rule for its own sake — a habit that forces you to keep deciding. If you can’t find a sentence worth keeping as-is, the tool has taken over, and it’s time to write the next paragraph with the tool closed.

Idioms: don’t borrow what you wouldn’t say

AI tools love idioms. Ask one to make your writing “more natural” and it will stuff in phrases like “hit the ground running” and “the ball is in your court” — phrases that sound natural coming from someone who grew up hearing them, and slightly costume-like coming from someone who didn’t.

Here’s my test: if you wouldn’t say the phrase out loud to a friend, don’t let the tool put it in your writing. Your readers can feel the difference between an idiom you own and one you rented. And a plain sentence you fully control — “We should start quickly” — beats a borrowed idiom every time. Plain English is not broken English. It’s the clearest English there is.

The same goes for slang and contractions the tool suggests. “Gonna” and “wanna” might be fine in a casual email, but if you don’t naturally use them, they’ll read as affectation. Write the English you actually speak. It’s better than the English you think you should speak.

A workflow that keeps your English yours

Here’s the routine I’d suggest, and it’s short enough to actually use:

First, draft in whatever language your thinking is clearest in — even if that’s a mix. Don’t perform in English while you’re still figuring out what to say.

Second, get the draft into English, with the tool’s help if you need it, and read it aloud. Your ear is your best editor here. If a sentence is hard to say, it’s hard to read, and it needs rewriting — by you, not by the tool.

Third, run the grammar checker, accept the fixes you understand, reject the ones that flatten your voice, and ask “why” about anything confusing.

Fourth, check the register: does this sound like the situation? Formal where it should be formal, plain where it should be plain? One read-through with that single question catches most tone mistakes.

Fifth, keep one sentence per paragraph the tool never touched.

That’s it. No special software, no premium plan, no ten-step system. Just you staying in charge of the decisions that make writing yours.

A quick checklist

• Drafted the idea in the language I think in, not the language I’m performing in
• Read the English version aloud — hard to say means hard to read
• Ran the grammar checker and accepted only the fixes I understand
• Asked “why” about at least one correction I didn’t expect
• Checked the register: formal or casual, matching the situation
• Rejected any idiom I wouldn’t say out loud
• Kept one sentence per paragraph the tool never touched

The honest truth

The goal was never to write English that hides where you’re from. The goal is to write English that carries what you mean — clearly, in your own voice, with your own ear for the language. AI tools can help you get there faster. They can also quietly replace your voice with a generic one while you thank them for it.

The difference is who decides. Keep deciding, sentence by sentence, and the tool stays a tool. That’s true for native speakers too — but for you, the stakes are higher, because your voice in English is something you built on purpose. It’s worth protecting.