Convert JSON to Excel (.xlsx)

The download is a genuine Office Open XML workbook — a ZIP archive of SpreadsheetML XML parts, the same container Excel itself writes — with a single worksheet named Sheet1. Row 1 is the header, built from the union of every object’s keys in the order they first appear: if the first record has id, name, salary, active, and skills, and a later record adds office, the sheet gets six columns and the earlier rows simply leave that cell empty. Missing keys never shift data sideways, because every row is written against the full header.

Cells are typed, not stringified. A JSON number becomes a numeric cell, so SUM, sorting, and filtering work the moment the file opens; true and false become real boolean cells that Excel displays as TRUE and FALSE; strings are written as inline text. That last point matters more than it sounds: an ID like "00501" stays text with its leading zeros intact instead of being coerced to the number 501, which is what CSV imports routinely do. null values and missing keys produce genuinely empty cells — the cell is omitted from the XML entirely, not filled with the word null.

Values that are themselves objects or arrays cannot be spread into a single cell without inventing a convention, so the converter keeps them lossless as compact JSON text: ["math","compilers"] lands in the cell exactly like that. Nothing is dropped, but if you want one column per nested field, flatten the JSON first — the flatten tool rewrites nested keys as dotted paths — and then convert the flat result. A single top-level object becomes a one-row sheet, and an array of plain values becomes one column named value, so the tool stays predictable whatever shape you paste.

CSV reaches Excel only through an import step, and the import step is where data gets damaged. Excel guesses the delimiter — and guesses differently in locales where the list separator is a semicolon — then coerces anything that looks numeric: leading zeros vanish from ZIP codes, long IDs get rewritten in scientific notation, and strings like 1/2 quietly become dates. A native .xlsx skips all of it. You double-click the file and every type is already decided, because the types are stored in the cells themselves.

Encoding is the other classic CSV failure. Excel tends to assume a legacy code page for .csv files unless a UTF-8 byte-order mark is present, so accented names and non-Latin text arrive as mojibake. SpreadsheetML is UTF-8 XML by specification, which means Zoë, Müller, and 東京 survive with no BOM tricks and no import dialog. If your data contains user-generated text in any language, this alone is a good reason to hand Excel a workbook rather than a text file.

CSV still wins when the consumer is not Excel: it streams, it diffs cleanly in git, and virtually every database loader, warehouse, and BI tool accepts it. If your destination is a bulk import or a data pipeline, use the JSON to CSV converter instead. This tool is for the case where a human will open the file in a spreadsheet and you want it to just work — no wizard, no cleanup, no explaining to a colleague which delimiter to pick.

The workbook is assembled entirely inside your browser tab: JavaScript serializes your rows to SpreadsheetML, computes the ZIP checksums, and concatenates the archive in memory. Clicking Download .xlsx saves those bytes directly from memory to disk. There is no upload endpoint anywhere in this flow — nothing to intercept, log, or retain — which is a stronger guarantee than a converter that uploads your file and promises to delete it after processing.

That matters because the JSON people push into spreadsheets is often exactly the JSON they should not paste into a random uploader: salary tables, customer exports, API responses that embed emails or tokens. Here the data stays on your device from paste to download. And if you want to inspect or reshape the JSON before converting — rename keys, prune fields, flatten nesting — the full editor works on the same JSON with formatting, validation, tree and mind-map views, and an AI copilot.

The output is deliberately minimal. You get one worksheet, always named Sheet1 — there is no option to split an array across multiple tabs. Cells contain values only: no formulas, no styles, no fonts or colors, no merged cells, no column widths, no freeze panes. If you need a formatted report, treat the download as clean raw data and apply the presentation in Excel afterwards.

Dates are the sharpest edge. JSON has no date type, so a value like "2026-07-21" is a string and is written as a text cell — it is not converted to an Excel date serial, and no number format is applied. Excel will not treat the column as dates until you convert it, for example with DATEVALUE or Text to Columns. The reverse asymmetry exists in the companion Excel to JSON tool, where genuine Excel dates come back as their underlying serial numbers.

Two smaller caveats. The ZIP entries are stored rather than deflate-compressed, so the file is somewhat larger than the same sheet saved from Excel — still perfectly valid, and resaving it from Excel compresses it. And because everything happens in browser memory, very large arrays are bounded by your tab’s RAM; this is a converter, not a streaming exporter. Empty arrays and objects with no keys are rejected with a clear error message instead of producing a blank workbook.