Fix garbled CSV text with the correct source encoding
Unreadable CSV text can result from decoding a file with the wrong character encoding. Keep the source and preview another encoding before exporting a readable copy.

Can this recover text that has already become question marks?
Choose the correct encoding when reading the original CSV bytes. If the characters were replaced with question marks and that version was saved, changing the encoding cannot restore them. Return to the original export, confirm the preview is readable, then save as UTF-8 with BOM.
How to use this tool
- Add the CSV and select its source encoding.
- Preview headers and Korean records for readable text, including names and quoted fields.
- Run the task to create UTF-8 output with BOM.
- Download the CSV and reopen it in the receiving program.
Choose the right settings
Select the encoding used by the source, considering EUC-KR/CP949 for Korean data. Confirm that headers and records are both readable before exporting the file as UTF-8 with BOM.
Before you start
Establish the meaning of the input
Before converting or arranging data, identify what a row or value represents and keep a source copy. In tabular files, review headers, separators and text encoding; in structured files, inspect the required object or array shape. Identifiers, dates and amounts should not be interpreted solely from how a spreadsheet happens to display them.
Where the tool offers matching columns, type rules or output fields, choose them according to the destination’s requirements. Test a sample with a blank, a repeated value and a value containing punctuation. For labels, charts or personalized documents, check which column supplies each displayed value and whether the source units and wording are correct.
Define a few checks before processing: an identifier that must retain its spelling, a value with a known interpretation and an expected relationship between records. If the tool offers type conversion or matching rules, distinguish a source value from the program’s display of it. Test ambiguous dates or differently written keys before accepting a broad rule. Keep the chosen rules and destination requirements with the result; this makes a discrepancy traceable to input, interpretation or the requested transformation.
Verify values, structure and exceptions
Reopen or import the result using the destination’s settings. Compare headers, record counts and several values with the source, allowing for the operation you requested. A summary, split or duplicate-removal task can legitimately change the count; verify why it changed rather than expecting every transformation to keep identical rows.
Inspect any exception or validation report before continuing. If columns shift, characters are garbled or a key is not found, check delimiters, encoding and exact header spelling first. Encoding data does not encrypt it, and formatting a value does not confirm its meaning. For printed codes or labels, test the actual output before making the full set.
Example input and output
EUC-KR/CP949 CSV → reviewed UTF-8 BOM CSVExamples describe the workflow. Results depend on the source file and settings.
Check the saved result before using it
Reopen the saved CSV in the program that will use it and inspect headers and names. Compare commas, quoted fields and row order with the source as well.
If the result is different from what you expected
If every encoding still produces damaged text, inspect the values in the original authoring program. Encoding conversion cannot recover lost characters or automatically reorganize the contents of the spreadsheet.
Limits to keep in mind
- Already corrupted characters cannot be recovered. Values, delimiters and row order remain unchanged; this resolves source-encoding problems.
Continue with a related tool
This guide explains the public browser tool. Keep a copy of the original and inspect the result before sharing it.