Grammar Checker: Inline Error Highlighting

You’re editing an article at midnight. The spell checker in your text editor catches a missing letter here and there, but the sentence that quietly switched tenses three paragraphs ago? It doesn’t notice. And the comma splice you’ve been writing your entire life? Invisible. A dedicated grammar checker works at a different level — it parses sentence structure, flags agreement errors, catches style issues, and does it across languages that your built-in spell check doesn’t even support.

This tool runs a full check against the LanguageTool engine. You paste text (up to 10,000 characters), select one of 25 supported languages, and get results displayed directly in your text with color-coded underlines. Every error is clickable, every suggestion is specific, and every fix is one click away.

Four Error Categories, Each With Its Own Color

Not all errors are equal, and this grammar checker treats them differently at a visual level. Grammar errors appear with red underlines — subject-verb agreement, tense inconsistencies, wrong prepositions. Spelling mistakes get amber. Style issues (redundancies, repetitions, overly complex phrasing) show up in purple. The color distinction matters because it lets you triage: fix the red grammar errors first, then decide whether the style suggestions are worth accepting.

The sidebar mirrors this system with filter tabs. Click “Grammar” and the list narrows to grammar-only issues. Click “All” to see everything. Each tab shows a badge count, so you know immediately whether your text has two spelling errors or twenty grammar ones. The summary panel at the top breaks this down further with labeled badges per category.

Every suggestion card in the sidebar shows the full error message from the engine, the original text with a strikethrough, and — this is where it diverges from simpler tools — all replacement options as clickable pills. The LanguageTool API often returns multiple possible corrections for a single error. Instead of silently picking the first one, this tool shows every option. The first suggestion is highlighted as the primary pill, but you can click any alternative. For a word like “there” used incorrectly, you might see “their” and “they’re” as separate pills. You pick the right one based on context.

What Happens When You Click Accept?

Behind the interface, the correction system works on a string-offset model. The API returns each error as an offset (character position) and length within the checked text. When you accept a fix, the tool replaces that exact substring in the working text, then recalculates every remaining error’s offset based on the length difference between the old word and the new one. This is why sequential fixes don’t break the text — each acceptance triggers a full offset recalculation and re-render.

This matters because grammar correction tools that manipulate the DOM directly (swapping HTML nodes instead of working with the underlying string) tend to produce garbled text after a few consecutive fixes. Words merge, spaces vanish, characters shift. The offset-based approach avoids that entirely.

The tool also includes undo and redo. Every accept, ignore, or “fix all” action saves a snapshot of the current text and error state. Press Ctrl+Z to step back, Ctrl+Y to step forward. The undo stack holds up to 50 states, which is enough for even aggressive editing sessions. “Fix All” applies the primary suggestion for every remaining error in a single pass, processing from the end of the text backward to preserve offset accuracy.

Twenty-Five Languages. Does It Actually Work Across All of Them?

The language selector includes English (US, UK, Australian), German (Germany and Austria), French, Spanish, Portuguese (Brazil and Portugal), Dutch, Polish, Italian, Russian, Ukrainian, Catalan, Danish, Greek, Romanian, Slovak, Slovenian, Swedish, Arabic, Japanese, and Chinese. There is also an auto-detect mode that identifies the language from the text itself.

Coverage quality varies by language. English, German, French, and Spanish have the deepest rule sets in the LanguageTool engine — thousands of grammar patterns, context-sensitive spelling, and style suggestions. Russian and Ukrainian have solid grammar coverage but fewer style rules. Japanese and Chinese support is more limited, focused primarily on common errors rather than nuanced style. The “Strict” checking level enables additional rules that the standard mode skips — useful for formal writing where even minor issues matter.

One practical detail about the proofreading tool worth noting: the checking level selector (“Standard” vs. “Strict”) isn’t just a label. Standard mode applies core grammar and spelling rules. Strict mode activates additional pattern checks — punctuation preferences, casing rules, typography standards, and more opinionated style suggestions. For casual text, standard is enough. For anything going into print or publication, strict catches things standard ignores.

The stats bar below the editor updates in real time as you type or edit: word count, character count, sentence count, and estimated reading time. After checking, you have two copy options — “Copy Text” grabs the current text as-is, and “Copy Corrected” applies all remaining first-choice suggestions and copies the result without modifying the editor. That second option is useful when you want to see what the fully corrected version looks like before committing. Your original text in the editor stays untouched until you explicitly accept fixes.

No account, no data transmission beyond the grammar API call. The text goes to the LanguageTool endpoint for analysis and the response comes back — that’s the only network request.