Catch the mechanical mistakes first.

What it checks

This is a punctuation checker first. It looks for the small mechanical problems that survive a fast draft: two spaces between words, a space before a comma, a missing space after a comma, an unmatched bracket, an odd number of quotation marks. Those are the errors that make a document look careless even when the argument is sound.

The current rule set covers:

• Spacing. Repeated horizontal spaces between non-space characters, a space before terminal or separating punctuation, and a missing space after a comma, semicolon, colon, exclamation mark, or question mark. Thousands separators such as 10,000and times such as 10:30 are left alone.
• Repeated words. A word repeated immediately, such as “the the”. The grammatical sequence “had had” and the common “that that” are allowed, because flagging them would create more noise than signal.
• Structure. Unmatched opening or closing parentheses, square brackets, and curly braces. Straight double quotes and curly quotation marks are also checked for balance.
• Sentence-level signals. A lowercase letter after a sentence-ending mark, a lowercase standalone “i”, and a sentence longer than forty words. These are hints, not verdicts.
• Common typos. A small list of frequent English misspellings — “recieve”, “seperate”, “definately”, and similar. The list is deliberately short and the checker does not correct these automatically.
• Repeated terminal marks. Runs such as !!, ???, or four or more dots. Three dots are treated as an ellipsis and are not flagged.

URLs and email addresses are masked before the rules run, so a comma inside a query string or an odd-looking domain is not mistaken for prose punctuation. That protection is not a license to paste code: source code, markup, and structured data contain patterns no prose checker should judge.

Safe fixes

The “Apply safe fixes” button changes only whitespace. It removes repeated spaces between words, removes a space before punctuation, and inserts a space after a separating or terminal mark when a letter follows. It does not change a single letter. That restraint is intentional: a rule that guesses at the right word creates a new error and hides the one you needed to see.

After a safe fix, the remaining issues stay in the list. An unmatched bracket still needs a human decision about where the closing bracket belongs. A misspelling still needs a human decision about whether it is a typo, a name, or a deliberate spelling. The button is there to clear the mechanical noise, not to finish the proofread.

Line-leading indentation is preserved. The repeated-space rule only acts between non-space characters, so a code sample or an indented outline is not flattened into a single column.

What it cannot do

It cannot check grammar. It does not know whether a verb agrees with its subject, whether a pronoun has a clear antecedent, whether a tense shift is intentional, or whether two clauses are joined by the right conjunction. It cannot see a comma splice unless a different rule happens to catch a symptom. It cannot tell you that a sentence is confusing, that a paragraph has no topic sentence, or that a claim is unsupported.

It also cannot replace a dictionary, a style guide, or a human reader. The typo list is tiny by design, the sentence splitter is heuristic, and the long-sentence threshold is a blunt instrument. A clean report means “the current rules found nothing”, not “this text is correct”. For legal, medical, academic, or published work, a careful human pass is still the last step.

Reading the score

The mechanical score starts at 100 and subtracts a weighted amount for each issue. Structural problems weigh more than style hints; a double space weighs less than an unmatched bracket. The result is useful for a quick before-and-after check when you are cleaning a batch of text, but it is not a percentage of correct sentences and it is not comparable across very different documents.

A 240-word cover letter with two spacing slips and one long sentence will score differently from a 40-word caption with a single unmatched bracket. That is expected. Read the issue list, not just the number. The category counts above the list tell you where the deductions came from: spacing, punctuation, typo, structure, or style.

Examples

The draft  is ready ,but the intro needs work.

The draft is ready, but the intro needs work.

The original has a repeated space and a missing space after the comma. Both are safe to fix without touching a word. A word-level checker might also suggest a stronger verb, but that is a different job and not something this tool pretends to do.

The summary is clear, but the appendix (which lists every source is still incomplete.

The summary is clear, but the appendix (which lists every source) is still incomplete.

The opening parenthesis has no closing partner. The safe-fix button cannot know whether the author meant to close it after “source” or delete the aside entirely. The checker reports the unmatched opening bracket and leaves the repair to you. That is the honest split between mechanical detection and editorial judgment.

Please recieve the attached file and confirm the date.

Flag: “recieve” → check “receive”
The rest of the sentence is left unchanged.

The checker can say that “recieve” is a common misspelling. It cannot know whether the word is part of a quoted error, a product name, or a code sample. Leaving the correction to the writer avoids a silent, wrong replacement.

A practical workflow

1. Apply the safe spacing fixes first. They are the least controversial changes.
2. Resolve structural flags — brackets and quotes — by reading the surrounding sentence.
3. Check each typo flag against the intended word. Do not accept a replacement blindly.
4. Split any sentence that is long enough to lose its subject. Read it aloud to hear the seam.
5. Run the checker once more, then read the text without the tool. The final pass is still yours.

The checker is fast because it is narrow. Use it for the narrow job, and use a grammar-aware editor or a human reader for the rest. A tool that promises everything usually hides what it is actually doing.

Limits

• The typo list is short and English-biased. It contains frequent misspellings, not a dictionary. Names, jargon, and deliberate spellings can resemble an entry, and many real misspellings are not on the list at all.
• Quote balance is not quote style. The checker can see an odd number of straight quotes or an unmatched curly quote. It cannot know whether the writer intended straight quotes, curly quotes, or an apostrophe that looks like a single quote.
• The sentence splitter is heuristic. Abbreviations, initials, decimals, and missing terminal punctuation can confuse the boundaries. A long-sentence flag may point at the wrong span, and a lowercase sentence-start flag can be a false positive.
• Safe fixes can collapse deliberate internal spacing. Line-leading indentation is preserved, but aligned columns inside a line are not. Do not run the fix on a table, a code sample, or any text where the spaces are structural.

Page updated 27 October 2026