Grammar Checker

Regular expressions are greedy about overlapping text, so the order of operations matters. Every pattern is run first and each match is recorded with its character offset; matches are then ordered by position (longest first on a tie) and scanned once, keeping a match only if it does not overlap something already kept. Without that step, 'I has a apple' would report one problem twice.

Sam is editing a short cover-letter paragraph before sending it: “He go to school every day. I has a apple. Their is a problem with alot of people. She don't know.”

1. Input:That is 21 words and 97 characters — a rounding error against the 5,000-character limit, and the rule engine re-runs on every keystroke, so the issue count updates as he types.
2. Grammar 1:“He go” is caught by the plain third-person pattern. A sibling rule deliberately ignores “go” when it is followed by “to” (to protect “He go to school” as a bare-infinitive reading), but the general rule still flags it and suggests “goes”.
3. Grammar 2 & 3:“I has a apple” produces two separate hits that do not overlap: “I has” → “have”, and “a apple” → “an” from the article rule. Because dedupe only drops overlapping matches, both survive.
4. Confusion:“Their is” → “there is”, and “alot” → “a lot”. A third pattern catches “She don't” → “doesn't”. The result is 6 issues in total: Grammar 3, Confusion 2, Articles 1.
5. Optional AI pass:Switching to “Rules + AI” and running the deep check downloads the local T5 model (about 335MB for Standard) and lists only the sentences it rewrites — useful for word-order and fluency problems no regex will ever catch.