# Date Format Normalizer Paste text with dates in mixed formats and get them all rewritten to ISO 8601 or any format you pick, with the day/month order settled from the text. Free and private. ## Run it - **CLI:** `gizza tool date-format-normalizer "Order 4471 was placed 03/04/2024 and shipped 15/04/2024. The invoice is dated 22 April 2024 and falls due 2024-05-06. Support ticket opened Friday, 3 May 2024 at 2:30 PM."` - **Web:** https://gizza.ai/tools/date-format-normalizer/ - **Agents:** machine-readable descriptor (parameters JSON Schema) at https://gizza.ai/tools/date-format-normalizer/tool.json ## Inputs - `text` — Text (dates can be in any mix of formats) _(field)_ - `output_format` — Rewrite every date as _(field)_ - `custom_format` — Custom pattern (only used by the custom format) _(field)_ - `separator` — Separator between the numeric fields _(field)_ - `month_style` — Month name written as _(field)_ - `year_style` — Year written as _(field)_ - `leading_zeros` — Pad single-digit days and months with a zero _(field)_ - `input_order` — Reading a numeric date like 03/04/2024 _(field)_ - `two_digit_year_pivot` — Two-digit years at or below this become 20xx _(field)_ - `keep_time` — Keep the clock time when a date has one _(field)_ - `time_style` — Clock time shown as _(field)_ - `output_timezone` — Move dates that carry an offset to (IANA name, UTC, or an offset) _(field)_ - `detect_timestamps` — Also read bare 10- and 13-digit numbers as unix timestamps _(field)_ - `output_mode` — Give me back _(field)_ ## Output - Normalized text (text) ## Query parameters Open the tool pre-filled and auto-run via URL: - `text` — Text (dates can be in any mix of formats) - `output_format` — Rewrite every date as - `custom_format` — Custom pattern (only used by the custom format) - `separator` — Separator between the numeric fields - `month_style` — Month name written as - `year_style` — Year written as - `leading_zeros` — Pad single-digit days and months with a zero - `input_order` — Reading a numeric date like 03/04/2024 - `two_digit_year_pivot` — Two-digit years at or below this become 20xx - `keep_time` — Keep the clock time when a date has one - `time_style` — Clock time shown as - `output_timezone` — Move dates that carry an offset to (IANA name, UTC, or an offset) - `detect_timestamps` — Also read bare 10- and 13-digit numbers as unix timestamps - `output_mode` — Give me back Example: `https://gizza.ai/tools/date-format-normalizer/?text=Order%204471%20was%20placed%2003%2F04%2F2024%20and%20shipped%2015%2F04%2F2024.%0AThe%20invoice%20is%20dated%2022%20April%202024%20and%20falls%20due%202024-05-06.%0ASupport%20ticket%20opened%20Friday%2C%203%20May%202024%20at%202%3A30%20PM.&output_format=iso&custom_format=%25d.%25m.%25Y&separator=dash&month_style=full&year_style=four&leading_zeros=true&input_order=auto&two_digit_year_pivot=68&keep_time=true&time_style=24h&output_timezone=source&detect_timestamps=true&output_mode=text` --- ## About this tool Real text almost never uses one date format. A document that passed through a spreadsheet, an export, an email thread and someone's handwriting ends up with `03/04/2024`, `15 April 2024`, `2024-04-22` and `5.1.24` sitting in the same paragraph. This tool finds every one of them and rewrites them into the single format you pick, leaving the rest of the text byte-for-byte untouched. The interesting part is not the rendering — it is deciding what `03/04/2024` means. Instead of guessing per date, the tool reads the whole text first and lets the dates that can only be one thing settle it for the rest: a single `15/04/2024` in the paste proves the source writes day-first, so `03/04/2024` is 3 April, not 4 March. When nothing in the text settles it, the month-first reading is used and the **report** mode tells you exactly which dates were guessed. ### Worked example Input: ``` Order 4471 was placed 03/04/2024 and shipped 15/04/2024. The invoice is dated 22 April 2024 and falls due 2024-05-06. Support ticket opened Friday, 3 May 2024 at 2:30 PM. ``` Output with the default ISO 8601 format: ``` Order 4471 was placed 2024-04-03 and shipped 2024-04-15. The invoice is dated 2024-04-22 and falls due 2024-05-06. Support ticket opened 2024-05-03T14:30. ``` Three source forms, one target form. `15/04/2024` settled the day-first reading for `03/04/2024`; the weekday name was absorbed into the date it belonged to; the clock time came along with its date; and the sentence's full stop survived. Switch the format to **Day first** with a **dot** separator and the same text renders `03.04.2024`, `15.04.2024`, `22.04.2024`, `06.05.2024`. Switch to **Year first** with the separator set to **none** and `scan 5 Jan 2024.pdf` becomes `scan 20240105.pdf` — the stamp that makes a folder of files sort itself. ### What it recognizes | Form | Examples | | --- | --- | | ISO 8601 | `2024-01-05`, `2024-1-5`, `2024-01-05T14:30:00Z`, `2024-01-05 14:30` | | Numeric | `01/05/2024`, `5.1.2024`, `5-1-24`, `2024/01/05` | | Month name | `January 5, 2024`, `Jan. 5th, 2024`, `5 Jan 2024`, `5th of January 2024` | | With a weekday | `Friday, 5 January 2024`, `Fri, 05 Jan 2024 14:30:00 +0100` | | Unix epoch | `1704465000`, `1704465000123` — only when you switch timestamp detection on | A clock time written next to a date is picked up with it, including the `at 2:30 PM` phrasing and a trailing `Z`, `UTC` or `+01:00`. ### Output formats ISO 8601 · year-first, day-first and month-first numeric (with your choice of dash, slash, dot, space or no separator) · full or abbreviated month names in either order · RFC 2822 email dates · Unix seconds or milliseconds · or any chrono/strftime pattern you want, such as `%B %-d, %Y` or `%Y%m%d`. Three ways to get the result back: the **text** with the dates rewritten in place, a **list** of just the normalized values one per line for pasting into a spreadsheet column, or a **report** giving each date's line and column, what was found, what it became, and whether the day/month order had to be guessed. ### Limits and edge cases - Up to **1,000,000 bytes** of text per run. Split anything larger and run the parts separately. - Strings that look like dates but are not — `2024-02-30`, `13/13/2024`, a version number like `1.2.2024` where no valid reading exists — are left exactly as written, never guessed at. - Bare 10- and 13-digit numbers are only read as timestamps when you ask for it, and then only when they land between 1973 and 2100. Order numbers and phone numbers stay untouched. - Two-digit years need a century: the pivot decides it, defaulting to the POSIX rule where 68 and below mean 20xx. - The timezone setting moves only the dates that carry an explicit offset. A date written with no zone has nothing to convert from, so it stays where it is. - Everything runs in your browser. No upload, no account, no network round-trip, and the same input always produces the same output. ## FAQ
How does it decide whether 03/04/2024 is 3 April or 4 March? It reads the whole text before rewriting anything. Any numeric date with a field above 12 can only be read one way — `15/04/2024` must be day-first, `04/15/2024` must be month-first — and those dates set the reading for every ambiguous date in the same text. If the text contains no such date, or contradicts itself, the month-first reading is used and the **report** output mode flags every date that was guessed, so you can check them instead of trusting them. You can also skip the inference entirely by setting the reading to **Day first** or **Month first** yourself.
Does it change anything other than the dates? No. In **text** mode the output is the input with each detected date span swapped for its rewritten form; every other character, including line breaks and punctuation, is passed through untouched. Two things are deliberately absorbed into the date they belong to: a weekday name written in front of it (`Friday, 5 January 2024` becomes one date) and a clock time written after it. A sentence-ending full stop after `2:30 pm.` is kept — only the `pm` is treated as part of the time.
What happens to times and timezones? A time found next to a date travels with it and is rendered on the 24-hour or 12-hour clock, with seconds shown only when the source had them. Turn **Keep the clock time** off to reduce everything to bare dates. The timezone field moves the dates that carry an explicit offset — an ISO stamp ending in `Z` or `+01:00`, an RFC 2822 date, a detected epoch value — into UTC, an IANA zone such as `Europe/Berlin` with daylight saving applied per date, or a fixed offset. That can change the calendar day, which is the point. Dates written without any zone are left exactly where they are.
Can I get a format that is not in the dropdown? Yes — choose the **custom** format and write a strftime pattern. `%d.%m.%Y` gives `05.01.2024`, `%B %-d, %Y` gives `January 5, 2024`, `%Y%m%d` gives `20240105`, `%A %d %b %y` gives `Friday 05 Jan 24`. The common fields are `%Y`/`%y` for the year, `%m`/`%d` for month and day (`%-m`/`%-d` without the leading zero), `%B`/`%b` for month names, `%A` for the weekday, `%H:%M:%S` for a 24-hour clock, `%I:%M %p` for a 12-hour one and `%z` for the offset. An invalid pattern is reported rather than half-rendered.
Why are my order numbers not being converted? Because timestamp detection is off by default. Most long numbers in real text are order ids, account references or phone numbers, and turning every 10-digit number into a date would ruin more documents than it fixes. Switch **Also read bare 10- and 13-digit numbers as unix timestamps** on when you are working with log exports or API payloads where epoch values really are dates — and note that even then only values between 1973 and 2100 are accepted.
How do I check what it actually did? Set the output to **report**. You get a `#` header with how many dates were found, the mix of source forms detected, and which day/month order was chosen and why, followed by one tab-separated line per date: its line and column in the source, the original string, the rewritten value, and an `ambiguous` marker when the day/month order had to be inferred. It pastes straight into a spreadsheet if you want to review a large document column by column.
## Related tools - [Absolute value, sign, or negation for a whole column](https://gizza.ai/tools/absolute-value-transformer/): Paste a column of numbers and apply absolute value, sign extraction (-1/0/1), sign flipping, or force-negative to every value at once, with rounding and an audit table. - [Adjacency Matrix Converter](https://gizza.ai/tools/adjacency-matrix-converter/): Convert a graph between edge list, adjacency matrix, and incidence matrix — directed or undirected, weighted or not. Free, private, runs in your browser. - [Amazon Order Analyzer](https://gizza.ai/tools/amazon-order-analyzer/): Paste an Amazon order-history CSV export to summarize total spend by month, top items, and category breakdowns. Browser-only, private, with Markdown or JSON output. - [ARFF Converter](https://gizza.ai/tools/arff-converter/): Convert Weka ARFF datasets to CSV and CSV tables back to ARFF locally — nominal attributes, numeric types, dates, sparse rows, missing values, and type rows. - [Avro to JSON Converter](https://gizza.ai/tools/avro-to-json/): Decode Apache Avro Object Container Files (.avro / OCF) to JSON, NDJSON, or the embedded schema — no .avsc needed, free and private in your browser.