Chinese CSV Files in Excel: UTF-8, GBK and Garbled Text

If a Chinese CSV opens as unreadable symbols, first keep an unchanged copy. The file may contain valid text bytes that Excel has interpreted using the wrong character encoding. Saving the garbled display back over the original can make recovery harder.

A CSV file stores text and separators. Its filename does not reliably identify its encoding, and it does not carry the same worksheet formatting as an Excel workbook. Reading the bytes correctly is the first step; preserving codes and dates in the spreadsheet is a separate step.

Choose the encoding used to save the file

UTF-8 is widely used for modern text exchange. Older mainland exports may use GBK or a related Chinese encoding, while some traditional-Chinese files use Big5. A traditional-looking phrase is not by itself proof that the file uses Big5; Unicode files can contain either script.

Ask the sender which encoding their export uses when possible. If that is unavailable, preview candidate encodings and inspect real words, uncommon names and punctuation. A decoder accepting the bytes without an error does not prove that the selected encoding is correct.

Use Excel's text import preview

Instead of relying on a double-click, use Excel's text or CSV import flow. In current desktop versions, this is commonly under the Data tab through a Text/CSV import option. Choose the source encoding and delimiter, then inspect the preview before loading the table. Menu names can differ by version and language.

A UTF-8 byte-order mark, or BOM, can help Excel recognise UTF-8 when opening a file directly. It is an encoding signature at the start of the file, not part of the first column's name. Some downstream systems prefer UTF-8 without a BOM, so choose the output expected by the next application.

Convert a readable preview to UTF-8

1. Open a copy of up to 5 MB in the Chinese file encoding converter.
2. Select the actual source encoding and read the preview.
3. Check several Chinese names or descriptions rather than only the header row.
4. Export UTF-8, choosing a BOM when appropriate for the receiving application.
5. Reopen the exported file and compare representative rows with the original source.

Changing a filename extension or adding a BOM does not transform GBK bytes into UTF-8 bytes. Conversion must first decode the source text correctly, then encode that text in the destination format.

Protect leading zeros and long codes

Even after the Chinese text displays correctly, Excel may interpret a code as a number or a date. Set relevant columns to text during import. A value such as 001234 should remain six characters when it is an identifier; displaying it as 1234 loses information.

Width differences can also affect matching after import. If one file has Ａ１２３ and another has A123, use a deliberate width conversion on that field. This does not repair an encoding problem and should not be applied indiscriminately to prose.

Separate three common symptoms

These problems can occur together, but they have different causes. Choosing UTF-8 will not by itself split a semicolon-separated file into columns when the importer expects commas. Selecting a delimiter will not recover Chinese characters that were decoded incorrectly. Setting a column to text protects identifiers but does not establish the file's original encoding.

Inspect the preview in that order: readable text, correct field boundaries and suitable column types. Use a few rows containing Chinese text, a code with a leading zero and a field that includes punctuation. A header row made only of ASCII letters can look correct under several encodings, so it is not enough to confirm the rest of the file.

Keep a small comparison sample

For an illustrative test file, use three fields: a name, a code and a note. The name can be 李明, the code 001234 and the note a short Chinese phrase. This is test content rather than a real customer record. Keep the expected text beside the file so you can compare the same values after each import or export.

If a field contains the file's delimiter or an embedded line break, the CSV writer must quote it correctly. A line count is therefore not always a row count: a quoted cell can contain a newline. When checking an export, compare the actual parsed rows and columns rather than declaring that a file is complete because its number of visible lines looks plausible.

After converting to UTF-8, reopen the new file using an explicitly chosen UTF-8 setting and compare the sample values. Then test the import in the application that will actually receive it. A file can be correctly encoded while still using a delimiter or column order that the receiving workflow does not expect.

Choose an output for the next step

UTF-8 with a BOM is useful for some direct-open Excel workflows because the signature helps identify the encoding. UTF-8 without a BOM may be the expected format for another importer. Neither choice is universally preferable; use the format specified by the next application and test it with the same representative rows.

A BOM cannot repair text that was decoded incorrectly before export. If the preview already shows the wrong characters, stop before saving and correct the source-encoding choice. Adding a UTF-8 signature to those wrong characters would merely label a new file that still contains the wrong text.

If you need to retain formulas, multiple sheets, cell styles or rich spreadsheet structure, a CSV is not a substitute for a workbook. Keep the workbook as the working document when those features matter, and export a CSV for the particular data-transfer step. That distinction prevents a successful text conversion from being mistaken for preservation of every spreadsheet feature.

Recognise lost information

If the saved file already contains replacement characters or literal question marks where Chinese text used to be, selecting another encoding may not restore it. Return to an unchanged original or request a fresh export. A converter can reinterpret existing bytes; it cannot reconstruct text that was overwritten and is no longer present.