Chinese typo checker. Scans your text against a curated dictionary of common 易错词 / 形近字 confusions and flags matches with the correct form + a note. Browser-only — a bounded reference checker, NOT AI / contextual correction.
Chinese Typo Checker (Reference)
How to use
Paste or type Chinese text
Drop the sentence, paragraph, or article you want to check into the box. The tool runs entirely in your browser — no text is uploaded to any server.
Click "Check for typos"
The tool scans your text against a built-in dictionary of common 易错词 / 形近字 (look-alike) confusions. Any matched WRONG form is highlighted. Note: this is a reference cross-check, NOT an AI corrector — it only catches mistakes that are in the curated list.
Read the flags + correct forms
Each match shows a "wrong → correct" card with a short note explaining the right character (e.g. 「再接再励」 should be 「再接再厉」).
Browse the reference list
Click "Browse reference list" to see every curated entry as a side-by-side table for study and cross-checking. If nothing matches, you'll see "No known common typos found".
Chinese Typos: Why They Happen, and What This Tool Catches
The Chinese term 错别字 (cuò-bié-zì) actually covers two distinct problems. A 错字 is a character written with genuinely wrong strokes — a non-existent shape. A 别字 is the far more common kind: writing one real character where a different real character was meant. Chinese is unusually prone to this because it is full of homophones (characters that sound identical or near-identical, like 在/再, 以/已, 的/地/得) and look-alike characters (visually confusable forms like 辨/辩/辫, 己/已/巳). When typing fast and picking candidates from a pinyin or zhuyin IME, almost everyone slips occasionally — and certain idioms carry "fossilised" miswritings so common they look right (e.g. writing 再接再励 instead of the correct 再接再厉). This tool is a reference checker aimed squarely at those high-frequency, unambiguous 别字 and easily-confused words.
It is a reference cross-check, NOT an AI corrector
Please understand the boundary clearly: this is a bounded, dictionary-driven reference checker, not an artificial-intelligence correction engine. Its method is simple — it compares your text, character by character, against a hand-curated list of "commonly-miswritten forms," and whenever a listed wrong form appears, it flags it and shows the correct version. It cannot judge from context whether a particular 「的」 is right in this sentence, and it cannot discover any new error that is not already on the list. In other words, it deliberately does not handle context-dependent characters (such as how 的/地/得 or 在/再 should be chosen within a sentence); it only includes forms that are almost always wrong regardless of context. A "none found" result does not certify your text as error-free — it only means none of the tool's curated typos appeared.
"Checks against common known typos — not AI / contextual correction." That is this tool's design principle and its honest boundary.
How to get the most out of it
Treat it as a first-pass proofreading sieve and a quick-reference study sheet for tricky characters — not the final word. Run a draft through it and it will instantly surface classics like 迫不急待, 再接再励, and 世外桃园, sparing you the chore of checking every idiom by hand. Open "Browse reference list" and the whole curated set of high-frequency confusions becomes a study list you can scan at once. For everything that genuinely needs human judgement (punctuation, grammar, and which homophone fits the sentence), a person should still read the whole thing — or a professional proofreader should. All comparison runs locally in your browser; nothing is uploaded or stored, so even sensitive drafts are safe. The curated set is intentionally bounded and grows over time, but it will never claim to be exhaustive.
10 Facts about Chinese Typos (错别字)
错别字 is a compound term: 错字 means a non-existent (mis-stroked) character, while 别字 means writing one real character where a different one was intended. Most everyday "typos" are actually 别字.
The two biggest causes of 别字 are homophony (在/再, 以/已) and visual similarity (己/已/巳, 辨/辩/辫). Picking the wrong IME candidate is a prime source of homophone errors.
Idioms are a hotspot: 再接再厉→再接再励, 世外桃源→世外桃园, 迫不及待→迫不急待. These slips are so common they often pass as correct.
的/地/得 are the most famous context-dependent confusables: 的 after attributives (美丽的花), 地 after adverbials (慢慢地走), 得 before complements (跑得快). They need sentence-structure judgement, which this tool deliberately does not attempt.
This tool is a bounded reference checker: it only matches a built-in list of common wrong forms, does no contextual analysis, and cannot find anything off the list — the fundamental difference from AI correction.
必须 vs 必需 are routinely mixed up: the adverb 必须 modifies actions (必须完成), while 必需 is the nominal "indispensable" (生活必需品). One character apart, different parts of speech.
度 vs 渡 trip many up: time/spending uses 度 (度假, 欢度, 度过难关); crossing water uses 渡 (渡河, 渡轮). 渡假 is a common error — the correct form is 度假.
Simplified→Traditional conversion can itself create 别字: simplified 发 maps to both 髮 (hair) and 發 (emit). A naive one-to-one conversion gets it wrong. This tool's Traditional list is sense-aware.
Some correct forms are often suspected of being wrong — e.g. 川流不息 (川 = river, not 穿) and 美轮美奂 (轮 = grand, not 仑). The reference list includes these correct forms as clarifications.
Machine matching efficiently catches high-frequency fixed errors, but punctuation, grammar, and context-bound homophones still need a human. Best practice is "machine first pass + human read-through" — this tool only does the former.
Frequently Asked Questions
-
No. It is a bounded reference checker — "checks against common known typos, not AI / contextual correction." It only compares your text to a built-in list of common wrong forms and flags matches. It does not analyse context or meaning, and cannot find errors that are off the list. For contextual, comprehensive correction you need a different (AI) tool — not this one.
-
No. Whether 的/地/得 is correct depends on sentence structure (的 for attributives, 地 for adverbials, 得 for complements) and requires contextual analysis. This tool deliberately includes only forms that are almost always wrong regardless of context, and skips context-dependent characters to avoid false positives. A human still needs to check these.
-
No. "No known common typos found" only means none of the tool's curated errors appeared — it does not certify the whole text. Anything off the list, plus punctuation, grammar, and context-bound homophones, is invisible to it. Read "none found" as "the first-pass screen found no obvious high-frequency 别字," not "proofreading passed."
-
It currently holds roughly a hundred-plus hand-curated common confusions and idiom miswritings. We only include entries that are unambiguous and almost always wrong — quality over quantity. The list grows over time, but the tool stays a bounded reference and never claims to be exhaustive.
-
No. All comparison runs locally in your browser with JavaScript — there are no server calls, your text never leaves your device, and nothing is logged. Even sensitive drafts are safe to paste in.
-
Because some correct forms are widely doubted (e.g. people suspect 川流不息 should be 穿流不息). These entries are included as clarifications — to confirm the right character. They appear only as reference in the browse list and are never flagged as errors during a scan.
-
Yes. The UI and list follow the current language: a Simplified UI uses the Simplified list, a Traditional (zh-TW) UI uses the Traditional list. The Traditional list is sense-checked character by character (e.g. distinguishing 髮 vs 發 for simplified 发), not a blind conversion, so the conversion itself does not introduce new 别字.
-
No. It is a first-pass sieve that efficiently catches high-frequency fixed 别字, but it cannot handle punctuation, grammar, context-bound homophones, or logical flow — things that need understanding. For publication or important documents, a human should still read through, or a professional proofreader should check. Best used as "machine first pass + human final review."
-
If the same wrong form appears several times, each occurrence is listed separately. In the highlighted preview, when two matched spans overlap the earlier one is kept to avoid double-marking the same text. Results are ordered by first appearance in the text — fully deterministic and reproducible.
-
Because that error is not in the built-in list. This is a bounded reference checker — it only finds curated common forms; any unlisted error, rare 别字, or context-dependent mistake will be missed. That is by design, not a bug. A miss is a candidate for expanding the list — feel free to send it our way.
Related News
You may be interested in these recent stories from our newsroom.
-
AWS Commits $1 Billion to Embedded AI Engineers as the Enterprise Fight Shifts to Deployment
AWS has committed US$1 billion to a Forward Deployed Engineering organisation that embeds its engineers inside customer teams, following pri...
-
ByteDance Launches Seedream 5.0 Pro, an Image Model That Outputs Editable Layers
ByteDance has launched Seedream 5.0 Pro, a professional image-generation and editing model that breaks a single render into more than ten ed...
-
Mira Murati's Thinking Machines Releases Inkling, Its First Open-Weight Model
Mira Murati's Thinking Machines has released Inkling, its first model: an open-weight, multimodal Mixture-of-Experts system with 975 billion...