How can I preserve Unicode characters when writing JSON in Python?

The character has not been lost. Python’s JSON encoder uses ensure_ascii=True by default, so it represents non-ASCII characters with \u escapes. When decoded as JSON, caf\u00e9 and café represent the same string. [S1]

If you want the file itself to show the readable character, pass ensure_ascii=False to json.dump(). Also explicitly open the file with UTF-8 encoding: json.dump() writes text to a text stream, and the file’s encoding determines how that text is stored. [S1][S2] The option changes how non-ASCII characters are represented in the JSON text; it does not change the string value you supplied.

Here’s a small example using the same sample data for both files. The first file uses the default escaping; the second writes non-ASCII characters directly. Both are opened explicitly as UTF-8, and each is loaded back and compared with the original object. [S1]

import json

original = {"greeting": "café — こんにちは"}

for filename, use_ascii in (
    ("escaped.json", True),
    ("readable.json", False),
):
    with open(filename, "w", encoding="utf-8") as f:
        json.dump(original, f, ensure_ascii=use_ascii)

    with open(filename, "r", encoding="utf-8") as f:
        loaded = json.load(f)

    assert loaded == original

In escaped.json, the accented and Japanese characters are written using Unicode escape sequences. In readable.json, they appear directly. The assert checks that loading either file reconstructs the original Python object; it does not require the file’s characters to look a particular way. You can open the files in a text editor to inspect the different representations, but the comparison is the check that matters for the decoded data.

First check that the data you passed to json.dump() contains the expected text, and that you opened the file with encoding='utf-8'. Use the same explicit encoding when reopening it. Then load the file and compare the relevant value—or the whole object—with the original, as in the example. [S1][S2]

If the loaded value matches but another program displays question marks or odd symbols, this example does not establish what that program expects or how it interprets the file. Check that program’s documented encoding settings. The Python documentation explains Python’s JSON serialization and UTF-8 file handling; it does not guarantee how every other application displays the output. [S1][S2] Also remember that ensure_ascii=False does not disable all JSON escaping: required syntax and control characters are still escaped. [S1]