What Is a Grammar Checker? Definition and How It Works

A grammar checker is software that identifies grammatical errors in text - incorrect verb tense, agreement errors, misplaced modifiers, run-on sentences - and suggests corrections. Unlike spell check, which only catches misspelled words, grammar checkers analyse sentence structure. On Mac, built-in grammar checking is limited to native text areas and catches only basic errors. Studies show humans reading text with grammar errors are 35% more likely to distrust the author, making grammar correction a practical professional concern.

How does grammar checking work?

Grammar checking uses two main technical approaches, often combined: rule-based systems and machine learning models.

Rule-based grammar checking encodes grammatical rules as patterns. The system parses each sentence, assigns a syntactic role to each word (subject, verb, object, modifier), and checks whether the relationships between them conform to known rules. A rule might state: "when a singular noun is the subject, the verb must be singular." The checker finds cases where a plural verb follows a singular subject and flags the mismatch.

Rule-based systems are interpretable and fast, but they require manual encoding of every error pattern. Languages are complex, and edge cases multiply quickly. A rule that catches "he run" will also flag valid constructions if written naively. Building comprehensive rule sets is labour-intensive, and the systems tend to produce false positives on informal or unconventional writing.

ML-based grammar checking trains a model on large corpora of text labelled with grammatical errors and corrections. The model learns statistical patterns that correspond to errors, generalising beyond the specific cases in its training data. This approach handles novel phrasings more gracefully and adapts better to informal registers. Research indicates ML-based grammar checkers catch approximately 60-70% of grammatical errors in standard professional text, with rule-based systems alone typically performing below 50%.

Both approaches depend on dependency parsing: building a tree structure that maps the grammatical relationships between words in a sentence. Without parsing the sentence structure, it is impossible to determine whether subject and verb agree, because you first have to identify which word is the subject and which is the verb.

What is the difference between grammar check and spell check?

Spell check and grammar check operate at different levels of language and catch different classes of error.

Spell check works at the word level. It compares each word in the text against a dictionary and flags words that do not appear. It does not understand sentences - a spell checker cannot distinguish "their" from "there" or "effect" from "affect", because both spellings are valid dictionary entries. Spell check is a necessary baseline but not sufficient on its own.

Grammar check works at the sentence level. It analyses the structural relationships between words to catch errors that involve correctly spelled words used incorrectly. Subject-verb disagreement ("the results shows"), incorrect tense ("yesterday I go to the meeting"), missing articles ("I have question"), and comma splices are all grammar errors that pass spell check invisibly.

A useful rule of thumb: if you could catch the error by reading each word in isolation, spell check will find it. If you need to understand the sentence structure to see the error, only grammar check will find it.

What types of errors does grammar checking catch?

Grammar checkers target a consistent set of error categories in professional writing:

Subject-verb agreement: The verb must match the number of the subject. "The team are ready" (British English) versus "The team is ready" (American English) is a legitimate variation; "the teams is ready" is an error.

Tense consistency: A narrative or argument that shifts tense without reason reads as an error. Grammar checkers flag mid-sentence or mid-paragraph tense shifts that appear unintentional.

Article errors: Non-native speakers in particular struggle with when to use "a", "an", or "the". Grammar checkers trained on large corpora can flag the most common article errors - "I have a important question" (should be "an") - reliably.

Punctuation: Run-on sentences (two independent clauses joined without appropriate punctuation), comma splices (two independent clauses joined with only a comma), and missing terminal punctuation are all grammar-level errors rather than spelling errors.

Sentence fragments: A dependent clause or phrase presented as a standalone sentence. In formal writing, this is an error; in informal or literary writing, it may be intentional. Good grammar checkers flag these without overriding the writer's intent.

What are the limitations of grammar checkers?

Grammar checkers miss a significant portion of errors, and their false-positive rate is high enough to be frustrating if the system is too aggressive.

The most significant limitation is contextual understanding. Grammar checkers analyse sentences largely in isolation. If the error only becomes apparent by understanding the broader argument of a document, the checker will miss it. "The policy increased efficiency" is grammatically correct - but if the preceding paragraph established that the policy decreased efficiency, the word "increased" is wrong. No grammar checker can catch that.

Grammar checkers also struggle with intentional style. Sentence fragments, unconventional punctuation, and informal register are all valid in the right context. A system calibrated for formal prose will over-flag creative or conversational writing.

Domain conventions are another gap. Legal writing, scientific papers, and technical documentation all have conventions that diverge from general grammar rules. A grammar checker trained on general text may flag correctly formatted legal citations or technical specifications as errors.

On Mac specifically, built-in grammar checking via NSSpellChecker only works in native apps. Electron-based apps including Slack, VS Code, and Discord bypass the framework entirely, leaving writers without any grammar correction in the tools they use most.

Frequently asked questions

What is a grammar checker?

A grammar checker is software that analyses sentence structure to identify grammatical errors - subject-verb disagreement, tense inconsistency, punctuation mistakes, run-on sentences - and suggests corrections. It operates at the sentence level, unlike spell check, which only compares individual words against a dictionary.

Is grammar check the same as spell check?

No. Spell check flags words not in the dictionary. Grammar check analyses sentence structure to catch errors involving correctly spelled words used incorrectly. You can have error-free spelling with serious grammar mistakes, and vice versa. Most writing tools run both simultaneously.

What grammar errors can checkers not catch?

Grammar checkers reliably miss contextual errors (where grammar is correct but meaning is wrong), intentional stylistic choices (fragments, informal register), domain-specific conventions, and errors that only make sense in the context of the broader document rather than the individual sentence.

Does Mac have a built-in grammar checker?

macOS provides basic grammar checking in native apps via NSSpellChecker. It catches simple structural errors but misses many patterns. It does not work in Electron-based apps like Slack, VS Code, or Discord, which bypass NSSpellChecker entirely.

Which grammar checker works in every Mac app?

Charm's Polish feature uses the Accessibility API to provide grammar correction system-wide, including in Slack, VS Code, and Discord. Polish fires at sentence boundaries, highlights corrections with a blue glow, and supports optional OpenAI API integration for enhanced accuracy.