Automatic conversion is fast. It is also literal. The engine replaces characters (and sometimes words) from dictionaries. It does not know your client, your genealogy, or your brand guidelines. These are the problems people hit most often after they press Convert.
None of them mean "never use a converter." They mean: know the failure modes, then proofread the places those modes collect.
One Simplified form, several Traditional forms
后, 发, 面 and 干 are famous examples. In Simplified Chinese one shape covers meanings that Traditional Chinese still splits. The converter picks from dictionaries and a short context window. It can still choose the wrong Traditional character in a name, a classical quote, or a one-word table cell.
This is the most important linguistic error, because the output looks fluent. A reader may not notice that a personal name has been rewritten into a common adverb. See one Simplified character, several Traditional forms for examples and a search list.
The other direction, Traditional to Simplified, hides the split instead of inventing it. That feels cleaner and is easier to miss as a problem. The cost appears later, when someone converts the Simplified file back and cannot recover the original Traditional distinctions.
The file was already broken
If the source is mojibake — æ soup, å salad, or boxes instead of characters — conversion multiplies the mess. OpenCC will happily map garbage to different garbage if the bytes happen to look like characters.
Open the source in Word or a text editor first. If you cannot copy a readable Chinese sentence, stop. Fix garbled text and confirm a real Unicode font before you convert.
A related trap is a PDF that only looks like text. If you cannot highlight a sentence, the page is a picture. Conversion will extract little or nothing useful. That is an OCR job, not an OpenCC job.
Mixed Japanese or Korean
Kanji and hanja look related to Chinese characters. They are not a drop-in OpenCC input. Japanese shinjitai simplifications followed a different history from mainland Simplified Chinese. A Japanese paragraph run through a Chinese converter can damage particles, kana-adjacent kanji, and words that only exist in Japanese.
If a document mixes Chinese and Japanese — a conference program, a bilingual manual, a name list — convert only the Chinese portions, or split the file. The same caution applies to Korean text that includes hanja. OpenCC conversion types are Chinese standards, not a general CJK rewriter.
Regional words and the wrong target
A Taiwan phrase table will not make Hong Kong readers feel at home, and the reverse is also true. The output may be readable and still "off," the way a British reader notices American truck and apartment.
This is not always a bug. It is a wrong conversion type. Choose Taiwan or Hong Kong on purpose. If you needed only characters and the wording became surprisingly local, you may have picked a phrase type by accident.
Names, brands and protected strings
Converters do not know your do-not-translate list. A company name, a drug name, a temple name, or a romanized brand that happens to contain Chinese characters can change. So can a person's given name that uses a rare character.
Before a large job, collect the strings that must stay put. After conversion, search for them. If a string must remain Simplified inside a Traditional document, or the other way around, leave it out of the automated pass and paste it back.
Layout, wrapping and fonts
Traditional characters are often wider. A slide title, a button, or a spreadsheet column that barely fit Simplified labels may overflow, wrap, or collide with a logo. That is a layout check, not always a conversion bug.
Fonts cause a second visual false alarm. If the output font lacks Traditional glyphs, you will see boxes or fallback shapes even when the code points are correct. Install or embed a font that covers the target standard. Details are in Chinese fonts and Unicode.
Excel has its own variant: formulas should stay formulas. If a label inside a formula was Chinese, it may convert. Cell values that look like numbers should not. If a sheet looks empty after conversion, look at fonts and at whether the source text was stored as text at all.
Hidden parts of the file
Word stores text in headers, footers, comments, footnotes, text boxes and tables. PowerPoint stores it in notes and masters. Excel stores it in comments, sheet names and defined names. A converter that only touched the body would leave a bilingual mess.
ChineseDocConverter is built to read those parts of the office formats it supports. You should still open the download and click through headers, notes and sheet tabs. Hidden text is where leftover Simplified or Traditional characters hide.
A repair order that saves time
- Confirm the source is readable Chinese, not mojibake or a scan.
- Confirm you picked the right regional type.
- Convert a sample, then the full file.
- Search merger characters and protected names.
- Check layout and fonts.
- Only then send the file.
If you convert, notice a problem, and convert again from the already-converted file, you make diagnosis harder. Go back to the original.
Frequently asked questions
The output looks fluent. Can I skip proofreading? Fluent is exactly how a wrong 後 in a name looks. Skim names and titles at minimum.
Why did only part of my PDF change? The rest is probably outlines, images, or a font subset that did not extract as text.
Can I undo a bad conversion? Only if you kept the original file. Do not rely on converting back.
Does a longer document fail more often? It fails in more places, not at a higher rate per sentence. Sample pages still predict most dictionary issues. They do not predict every name.