JSON to CSV Converter

Turn API responses, database exports and JSON Lines into a spreadsheet. Nested fields become columns, and the file opens cleanly in Excel. Your data never leaves the browser.

How to convert JSON to CSV

1. Paste JSON into the text box, or drop .json, .jsonl or .ndjson files onto the box above — several at a time is fine. Pasted JSON is converted as you type.
2. Check the preview table. If the JSON holds more than one list, choose the one you want under Rows from. Change Nested objects, Arrays or Delimiter if needed; everything is converted again straight away.
3. Click Download CSV, or Download Excel for an .xlsx file. Copy CSV puts the text on your clipboard, and with two or more files Download all (ZIP) saves every CSV at once.

How nested JSON becomes columns

A spreadsheet is a flat grid and JSON is not, so every value needs a column to live in. The converter reads all the records before writing anything and builds the header from every key it finds, not just the keys of the first record. A key inside another object becomes a dotted path, so address.city gets a column of its own. Columns appear in the order they are first seen, grouped under their parent: a field that turns up only in a later record still lands next to its siblings instead of at the far right. A record that lacks a field gets an empty cell.

For example, [{"id": 1, "name": "Ana", "address": {"city": "Lisbon"}, "tags": ["vip", "beta"]}, {"id": 2, "name": "Ben", "address": {"city": "Leeds", "zip": "LS1"}}] gives five columns: id, name, address.city, address.zip and tags. Ana’s zip and Ben’s tags are empty, and Ana’s tags read vip; beta.

Lists of plain values are joined with a semicolon and a space. A list that contains objects, such as the line items of an order, is kept as compact JSON in a single cell; spreading it out would create items.0.sku, items.1.sku… columns that differ from row to row. To keep the original structure, set Nested objects to JSON text: the top-level keys stay as columns and each nested object is written as JSON in its cell. Set Arrays to JSON text as well and lists of plain values are kept as JSON too, instead of being joined.

Opening the CSV in Excel without garbled characters

A CSV file is plain text with nothing inside that says which character encoding it uses. When you double-click one, Excel for Windows often falls back to the computer’s older regional encoding unless the file starts with a byte order mark (BOM), and UTF-8 text such as café comes out as café. The UTF-8 BOM option, on by default, writes the three bytes EF BB BF at the start of the file so Excel reads it as UTF-8. Copied text never includes the BOM.

Turn the BOM off when the CSV is meant for a program rather than a person, because some import scripts read it as part of the first column name. For a CSV someone else made without a BOM, use Data › From Text/CSV in Excel and set File Origin to 65001: Unicode (UTF-8). Or skip the question and download the .xlsx, where text is stored as Unicode and there is no encoding to guess.

JSON shapes that convert

You don’t have to tell the converter how your JSON is laid out. It looks at the structure, picks a sensible starting point and shows the result in the preview straight away. These are the layouts it recognises:

Pasted text or a file that fails to parse as one JSON document is tried as JSON Lines when its first non-empty line parses on its own as an object or array and there are at least two non-empty lines; Windows line endings and blank lines are fine. Files named .jsonl or .ndjson are read line by line first and, if that fails, as ordinary JSON, which helps when a pretty-printed array was saved with the wrong extension.

Choosing where the rows come from

API responses rarely start with the records themselves. They wrap them: {"meta": {"page": 1}, "data": […], "links": […]}. The converter looks a few levels into the object for lists of records, and for objects whose values are records keyed by an ID. Whenever the top level is an object, a Rows from menu appears above the preview with every candidate and its row count.

The default is the largest list stored under a familiar name — data, items, results, records, rows, hits, features and a few more — and otherwise the candidate with the most rows. A document that is itself keyed by IDs gets one row per key. The menu also offers:

• Each top-level key as a row — shown when at least half of the top-level values are objects, such as a settings file with one block per environment.
• Whole JSON as one row — every field of the object becomes a column of a single row.
• One row per field (field, value) — a two-column table with the dotted path on the left and the value on the right. It suits translation files and long configuration objects that would otherwise turn into one very wide row.

Long numbers, leading zeros and dates

Excel stores numbers with 15 significant digits, and when it opens a CSV it converts anything that looks like a number. A 19-digit ID such as 1453928392010284033 is kept as 1453928392010280000 and shown in scientific notation, like 1.45E+18; once the file is saved again, the lost digits are gone. Leading zeros disappear as well (the ZIP code 02134 becomes 2134), and a phone number written as +14155550123 loses its plus sign.

The Keep long numbers & leading zeros option, on by default, protects three kinds of values in the CSV: whole numbers with 16 or more digits, digit strings that start with 0, and digit strings that start with +. Each is written as ="02134", a tiny formula that Excel displays as the original text. A note under the preview says how many cells were wrapped, and a red warning appears if the option is off and such values exist.

Two caveats. The ="…" trick is meant for Excel: a database import, a script or another spreadsheet app may read the formula text literally, so switch the option off when the CSV is going anywhere else. And the option does not touch date-like text such as 1-2, 3/4 or SEPT2. Excel’s habit of turning those into dates is the reason the HUGO Gene Nomenclature Committee renamed genes such as SEPT1 and MARCH1 (now SEPTIN1 and MARCHF1) in 2020. If your data has values like these, download the .xlsx: every JSON string is stored there as a text cell, and Excel leaves text cells alone.

Numbers with 12 to 15 digits, such as millisecond timestamps, keep their value, but Excel’s General format shows numbers of 12 or more digits in scientific notation. Select the column and apply the Number format with 0 decimal places to see every digit.

Comma, semicolon or tab

When Excel opens a CSV by double-click, it splits the columns with the list separator from the computer’s regional settings. That separator is a comma in the United States, the UK and China, but a semicolon in much of continental Europe, where the comma is the decimal mark. If the whole file lands in column A, choose Semicolon and download again. The decimal point stays a dot either way, so in those regions a value like 12.5 may still be read as text or as a date; the .xlsx avoids that, because numbers are stored as numbers and shown in the reader’s own format.

Tab produces a .tsv file and turns the buttons into Download TSV and Copy TSV. Tab-separated text is the easiest thing to paste: put it into Excel or Google Sheets and it spreads across columns by itself, with no import dialog.

What the CSV file contains

• The first line is the header, and every line ends with CRLF, the line break that RFC 4180 specifies — the last one included.
• A value is wrapped in double quotes when it contains the delimiter, a double quote, a line break, or spaces at either end. Quotes inside are doubled, so say "hi" is written as "say ""hi""".
• Numbers are written the way JavaScript prints them: 1.50 becomes 1.5 and 1E5 becomes 100000. Whole numbers too long for JavaScript, such as 64-bit IDs, are copied digit for digit from the source text instead of being rounded.
• true and false stay as words; null and missing fields become empty cells.
• Apart from the protected values above, text that starts with =, +, - or @ is written as it is. Spreadsheet apps may treat such a cell as a formula, so take care with CSV files built from data that strangers typed in, such as form submissions.

Size limits and speed

• Each file can be up to 200 MB. Pasted text has no fixed limit; anything over a million characters is kept out of the text box so the page stays responsive, and is converted from the list instead.
• The preview shows the first 100 rows and 30 columns, with a note when it is cut short. The downloads always contain everything.
• Copying works up to 10 million characters; above that, the copy button is greyed out and the download is the way to go.
• Excel holds at most 1,048,576 rows and 16,384 columns per sheet, and 32,767 characters per cell. The page warns when your data goes over: the .xlsx keeps what fits, and Excel itself stops at the same row limit when it opens a longer CSV. Very wide nested JSON can exceed the column limit; setting Nested objects to JSON text needs far fewer columns.
• In a test in desktop Chrome, a 25 MB file with 100,000 records converted in about half a second, and the matching .xlsx took about a second. Phones have much less memory, so very large files are better converted on a computer.

When the JSON won’t parse

Invalid input gets a message with the line and column of the first problem, a copy of that line and a ^ pointing at the spot. These are the usual causes:

• A comma before } or ] — easy to leave behind after deleting the last item by hand. Remove it.
• Single quotes, True, False or None — this is a Python dictionary printed with print(), not JSON. Output it with json.dumps(data) instead. NaN, Infinity and undefined are not JSON either; use null.
• Property names without quotes — a JavaScript object copied from source code. Run JSON.stringify(obj) in the browser console and paste the result.
• Comments — files such as tsconfig.json accept // comments, but standard JSON does not. Delete them first.
• The text ends too early — the copy was cut off, which happens with long console or log output. Copy the whole response again.
• Extra text after the JSON ends — two documents pasted back to back. Put each one on a single line of its own and the input is read as JSON Lines.

Other ways to convert JSON to CSV

• Excel: in Excel for Microsoft 365 on Windows, choose Data › Get Data › From File › From JSON. Power Query opens the file as a list; click To Table, then the expand button in the column header, once for each nested level, and load the result. It pays off when you import from the same source every week.
• Python: pandas.json_normalize(data) flattens nested keys into dotted column names, and .to_csv("out.csv", index=False, encoding="utf-8-sig") saves the result; the utf-8-sig encoding adds the BOM that Excel on Windows needs.
• Command line: jq’s @csv filter writes CSV lines, but you list the columns yourself, and nested objects have to be flattened first.

For a one-off file, or on a computer where you can’t install anything, the converter at the top of this page is quicker: paste, check the preview, download.

Frequently asked questions

More free tools