Convert CSV to JSON

How the CSV-to-JSON conversion works

CSV is flat, plain-text tabular data with no defined data types and no nesting. JSON is structured and can express both. Converting from one to the other means making a couple of real decisions, which this tool makes transparently rather than silently:

• Delimiter detection — the tool tests comma, semicolon, tab and pipe against the file's own first lines and uses whichever produces a consistent column count, since many CSV exports (especially from European-locale software) use a semicolon, not a comma.
• Header detection — if the first row looks like column names (mostly non-numeric, with the row below it containing numbers), it's used as the key for every JSON object. If it doesn't look like a header, the tool falls back to an array of arrays instead of guessing wrong keys.
• Quoted fields — a field wrapped in double quotes can safely contain commas, line breaks or an escaped quote (""), per RFC 4180. This is parsed with a proper state machine, not a naive split on the delimiter.

Why convert CSV to JSON

JSON is the format nearly every API, web app and modern database (document stores in particular) expects. If you exported a report or a spreadsheet as CSV and now need to feed it into code, an API request body, or a NoSQL database, converting it to an array of well-formed objects is usually the very next step — and doing it by hand for anything beyond a few rows is slow and error-prone, especially once quoted fields or inconsistent delimiters are involved.

Edge cases this tool handles

Semicolon-delimited CSV (common in Europe)

Detected automatically — no need to pre-convert the file or tell the tool which delimiter to expect.

Quoted fields containing commas or line breaks

Parsed correctly per RFC 4180, rather than being split into the wrong number of columns.

Missing values

A row with fewer values than the header still produces a valid JSON object — missing fields are simply empty strings, not null or omitted keys, so every object in the output has the same shape.

No header row

If the first row doesn't look like column names, the tool outputs an array of arrays rather than inventing or misusing column names.

Other ways to convert CSV to JSON

Neither Excel nor Google Sheets can save a sheet as JSON — JSON isn't in Excel's Save As list or in Sheets' File > Download menu, and Excel's Power Query can import JSON but not export it. So once you go beyond this page, the realistic options are a short script or a command-line tool. All of the commands below were run against a sample file named orders.csv.

Python, with nothing to install

Python's standard library can do it in one line. utf-8-sig silently strips the byte-order mark Excel adds to "CSV UTF-8" files, so your first key doesn't come out as order_id:

Python with pandas

pd.read_csv("orders.csv", dtype=str, keep_default_na=False).to_json("orders.json", orient="records", indent=2, force_ascii=False) produces the same array of objects. dtype=str keeps ZIP codes like 02134 intact, keep_default_na=False stops empty cells and the text "NA" from turning into null, and force_ascii=False writes "Zoë" instead of "Zoë". Leave out dtype=str if you want pandas to guess numeric columns for you.

PowerShell on Windows

Import-Csv .\orders.csv | ConvertTo-Json | Set-Content .\orders.json -Encoding utf8 works out of the box. Add -Delimiter ';' to Import-Csv for semicolon files. One quirk: a CSV with a single data row becomes a single JSON object rather than an array — in PowerShell 7 add -AsArray to ConvertTo-Json to always get an array.

csvkit on macOS or Linux

If you already use csvkit, csvjson orders.csv > orders.json converts in one step. Unlike this tool, csvkit guesses column types, so numbers and booleans come out unquoted; check ID-like columns afterwards.

Numbers, nulls and nested JSON

This converter writes every value as a JSON string — "price": "19.99", not "price": 19.99. That's deliberate: CSV has no data types, and guessing is how ZIP code 02134 becomes 2134 and a 17-digit order number quietly changes its last digits (JavaScript numbers can't hold integers above 9,007,199,254,740,991 exactly). If you need real numbers, convert only the columns you know are numeric. With jq:

• Make a column numeric — jq 'map(.price |= (tonumber? // null))' orders.json turns "19.99" into 19.99 and an empty price into null instead of failing.
• Empty strings to null everywhere — jq 'map(with_entries(if .value == "" then .value = null else . end))' orders.json.
• Nested objects from dotted headers — a CSV can't nest, but headers like customer.name and customer.city can be folded into "customer": {"name": …, "city": …} with jq 'map(reduce to_entries[] as $e ({}; setpath($e.key | split("."); $e.value)))' orders.json. This tool itself keeps the dotted names as flat keys.

Accented letters turn into � after converting

The tool reads the file as UTF-8, which is what Google Sheets, Numbers and Excel's "CSV UTF-8" option all write. Excel for Windows' plain CSV (Comma delimited) option instead saves in the old Windows code page, so letters like é or ñ arrive as invalid bytes and show up as the � replacement character. Re-save from Excel with File > Save As > CSV UTF-8 (Comma delimited) (*.csv) and convert again — the XLSX to CSV page explains the difference between the two options.

Related file extensions & tools

Frequently asked questions