Excel CSV Korean Text Corruption: The 3-Byte UTF-8 BOM Fix

Contents
This post is for anyone who opened a CSV file in Excel and found the Korean text replaced with strange Latin characters like 수출. Here's the short answer: the file isn't damaged. Excel just didn't read it as UTF-8.
A CSV file has nowhere to write "I am UTF-8." So you send the signal by putting the three BOM bytes at the very start of the file, and Excel reads the file as UTF-8 when those three bytes are there.
Summary ① The garbled characters are UTF-8 bytes read with a different character set. The bytes themselves are untouched, so you can get them back. ② The UTF-8 BOM is the three bytes
EF BB BFat the start of the file, and programs that read it correctly never show it on screen. ③ Without editing the file at all, you can open it in Excel through Text/CSV import on the Data tab, setting the encoding yourself instead of double-clicking.
Why you see characters like 수출
Save the Korean word for "exports" as UTF-8 and you get six bytes: EC 88 98 EC B6 9C. Each Korean character usually takes 3 bytes, and the word is two characters long.
When Excel doesn't recognize the file as UTF-8, it maps those six bytes straight into an old character set that pairs one byte with one character. In the Western European set (Windows-1252), EC is ì, 88 is ˆ, and 98 is ˜.
So the six bytes show up as six characters: 수출. Nothing disappeared. The same bytes were just read with a different table, which is why the Korean text looks fine when you open the same file in an editor that reads UTF-8.
The garbled shape depends on which table did the reading. If the file is read as the Korean code page (CP949), bytes get grouped in pairs, so order number can come out as a meaningless mix of Chinese characters and Hangul, like 二쇰Ц踰덊샇.
CSV has nowhere to record its encoding
RFC 4180, the spec that defines the CSV format, has no way to mark a character set inside the file. It only says that "commonly, US-ASCII is used, but other character sets may be used with the charset parameter."
That charset is a value in the transfer header attached when the file is sent. In other words, the encoding information lives outside the file.
The moment you download the file and save it, that header is gone. All that's left is a pile of bytes, so whatever opens the file has to guess.
What the three BOM bytes do
BOM stands for Byte Order Mark. It's the Unicode character U+FEFF placed at the start of the file. Encoded as UTF-8, that character becomes three bytes: EF BB BF.
Originally it announced which byte order an encoding with two possible orders, like UTF-16, was using. UTF-8 has only one byte order, so there's nothing to announce; here it works only as a signature that says "this file is UTF-8."
Those 3 bytes are not content. The web standard (WHATWG Encoding Standard) UTF-8 decode procedure says that if the first three bytes are EF BB BF, you discard them and treat the rest as content.
Microsoft's Excel support docs say the same thing. A CSV file encoded in UTF-8 opens normally if it was saved with a BOM; if it wasn't, they tell you to use Get Data on the Data tab.
How to fix it without a tool
There are three ways. Pick one based on whether you're allowed to change the file.
1. Import it in Excel instead of double-clicking
This method doesn't touch the file at all. On Excel's Data tab, choose Get Data → From File → From Text/CSV, then select the file.
A preview window opens with a File Origin list at the top. Choose 65001: Unicode (UTF-8) and the Korean text in the preview below immediately looks right.
The downside: if several people receive the file, each of them has to repeat these steps.
2. Re-save it with a BOM in Notepad
Open the file in Windows Notepad and go to File → Save As. At the bottom of the save dialog there's an Encoding list. Choose UTF-8 with BOM.
That's exactly why UTF-8 and UTF-8 with BOM are separate entries. Notepad's default these days is the version without a BOM, so saving as-is leaves the problem in place.
3. Pick the format when you export from Excel
Going the other way — exporting a sheet from Excel to CSV — choose CSV UTF-8 (Comma delimited) as the file type in Save As. That format saves with a BOM.
The problem comes when you already have a pile of downloaded files. Notepad struggles with large files, importing has to be repeated every time you open one, and choosing a file type doesn't help files that already exist.
Check it with our free tool, Excel CSV Character Fixer. Upload a file and it first shows a preview of the first 5 lines so you can see whether the Korean text reads correctly, then with one click you download a new CSV with a BOM at the front.
Add the BOM and download your file right in the browser, with no install and no sign-up. Uploaded files are never sent to a server.

Common sticking points
If the text is still garbled after you add a BOM, the file probably wasn't UTF-8 to begin with. A BOM only marks the file as UTF-8; it doesn't convert another encoding into UTF-8.
That's why the tool shows the first 5 lines first. The preview reads the file as UTF-8, so it gives you a way to tell.
| What the preview shows | What it means |
|---|---|
| Korean text looks normal | The file is UTF-8. Add a BOM and it opens in Excel |
Diamond question marks like � | The file isn't UTF-8 (CP949, for example). Don't download it |
| Korean mixed with Chinese characters and symbols | Same as above. You need to export the original again as UTF-8 |
If the preview looks broken and you download it anyway, the characters that couldn't be read get locked in as replacement characters and the originals can't be recovered. In that case the only fix is to have whoever created the file export it again as UTF-8.
There's a side effect on the other side, too. Sometimes, when a program reads the CSV, the first column name comes through as \ufeff stuck to the front of the name. That means the reader didn't strip the BOM.
Summary
- The garbled characters are the result of reading UTF-8 bytes with an old one-byte-per-character table. The bytes
EC 88 98 EC B6 9Cfor the Korean word for "exports" end up looking like수출. - CSV doesn't record its encoding inside the file. RFC 4180 treats
charsetas transfer-level information only. - The UTF-8 BOM is the three bytes
EF BB BFat the start of the file — the encoding of Unicode U+FEFF. - If you don't want to touch the file, go to Excel's Data tab → From Text/CSV and set File Origin to
65001: Unicode (UTF-8). - If you want to fix the file, save it as
UTF-8 with BOMin Notepad, use Excel'sCSV UTF-8 (Comma delimited)format, or use the tool below. - If adding a BOM doesn't help, the original isn't UTF-8. Check the preview first.
Add the BOM and download your file right in the browser, with no install and no sign-up. Uploaded files are never sent to a server.
Frequently Asked Questions
Does adding a BOM change the file's contents?
The text itself stays the same. Three bytes, EF BB BF, are added to the very beginning. The file size grows by 3 bytes.
Programs that follow the standard read those 3 bytes and discard them, then treat the rest as the body, so what you see on screen doesn't change. If you upload a file that already has a BOM, it won't be added twice. That's because the UTF-8 decoding procedure browsers use when reading a file strips those first 3 bytes first.
It looks fine in Notepad but breaks in Excel. Why?
The two programs decide on encoding differently. Notepad tries UTF-8 first, while Excel doesn't treat a CSV as UTF-8 when it opens one directly and there's no BOM. The same file gives different results not because of the file, but because the reader makes a different call.
I added a BOM and it's still broken.
That happens when the original isn't UTF-8. A BOM is only a label saying "this file is UTF-8." It doesn't convert bytes from another encoding. If the Korean text already looks broken in the tool's 5-line preview, this is that case, and you need to export the file as UTF-8 again from the system that created it.
Does the CSV file I upload get sent to a server?
No. Reading the file, previewing it, and creating a new file with a BOM all happen inside your browser, and the file never goes to a server. Even a CSV with internal order records or a client list doesn't leave your machine.
This article was adapted from our Korean original. Read the Korean original