CSV stores bytes, and those bytes need an interpretation
A delimiter tells a parser where fields separate. An encoding tells it which characters the bytes represent. A file can have perfectly consistent rows and still display damaged names or punctuation when read with the wrong encoding.
Common choices
| Encoding | What to check |
|---|---|
| UTF-8 | A useful default for exchange. Check accented letters, non-Latin text, and punctuation in the preview. |
| UTF-8 with BOM | A byte-order mark can help some software identify UTF-8. It should not become visible text in the first header name. |
| Windows-1252 | Appears in some legacy exports with Western European text. Selecting it for the wrong file can produce plausible but incorrect characters. |
| UTF-16 LE / BE | Byte order matters. Prefer an explicit source-system setting when it is available. |
Detection is a clue, not a guarantee
A short file containing only basic ASCII characters cannot demonstrate which of several compatible encodings its exporter intended. Even a longer file can be ambiguous. Use export documentation, a reliable BOM, and a review of representative text together.
If your source system says the file is Windows-1252, choose that setting. If you see text that resembles café instead of café, investigate decoding history; saving the damaged text as UTF-8 does not by itself restore the original bytes.
Review before you re-export
- Keep a copy of the original bytes.
- Select the expected encoding and delimiter.
- Inspect names, currency symbols, accented text, and the first header.
- Read encoding warnings and avoid silent replacement of undecodable characters.
- After downloading an output, open it using the encoding expected by your destination.
What TabularGate does
The validator exposes UTF-8, Windows-1252, UTF-16 little-endian, and UTF-16 big-endian choices. Auto-detection recognizes a UTF-16 byte-order mark and otherwise attempts UTF-8; choose Windows-1252 explicitly for a known legacy file. The report shows the interpretation used. File upload checks the actual bytes; pasted text has already been decoded by the application you copied it from.
Output serialization may change the file encoding even when decoded field values are unchanged. Use the preview and downloaded report to distinguish a character error from a harmless difference in file bytes.
Check your CSV encoding in the validator →
Further reading: the W3C tabular data model's encoding guidance recommends UTF-8 for publishing CSV on the web. Compatibility still depends on the receiving system.