Find and replace text in a CSV

How to find and replace in a CSV

1. 1. Open find & replace in your free workspace. Your file is read into the browser’s memory — never sent over the network.
2. 2. Pick the column and enter your text. Type what to find and what to replace it with; leave the replacement empty to simply delete the matches.
3. 3. Choose how to match. Optionally match case, match the entire cell only, and keep the original column alongside a new _cleaned one.
4. 4. Preview and download. See exactly which cells changed, then create a free account to export the cleaned CSV.

What it matches — and what it will not

Every match here is literal. The text you type is compared character for character, so *, ?, and . mean themselves. Excel takes the opposite approach: its search treats ? as any single character and * as any run of characters, which is powerful when the pattern varies and a hazard when it does not — a replace-all for 12.50 in a wildcard-aware search can reach 12X50 too. Use Excel for a genuine pattern, and use this for text you can name exactly.

Three settings decide how far a replacement travels:

• One column, not the sheet. The replacement is scoped to the column you pick, so swapping “NY” for “New York” in a state column cannot touch a company name in the next column over.
• Match entire cell only. Off, every occurrence inside the cell is replaced. On, the cell changes only when it equals the search text outright — the setting that stops “US” rewriting the middle of “Austin”.
• Match case. Off by default, so Ltd also finds LTD and ltd. Turn it on when the casing is the thing you are fixing.

Find & replace is the right tool for a known token — a stray Ltd., a currency symbol, a country spelled three ways. It is the wrong tool when the fix is structural. Address columns that need lowercasing and de-spacing belong in the email cleaner, repeated rows belong in the duplicate remover, and if the same swap has to run on a workbook instead, the Ctrl+H walkthrough for Excel covers the wildcard and scope options in full.

When find and replace says “0 replacements”

A search that finds nothing is almost never a broken search. It is usually that the cell does not hold what it appears to hold. Three causes account for most of it:

• An invisible character sits inside the value. Data pasted from a web page carries non-breaking spaces, which look identical to normal ones. Microsoft’s TRIM documentation is blunt about it: TRIM handles the 7-bit ASCII space (value 32), and “by itself, the TRIM function does not remove this nonbreaking space character” (value 160). So Acme Ltd can refuse to match Acme Ltd. Replace the invisible character first — search for the pasted space itself and replace it with a normal one — then run the real swap.
• Match entire cell is on when it should not be. With that setting on, Ltd matches a cell containing exactly Ltd and nothing else. If the cell reads Acme Ltd, it is skipped and the count comes back zero.
• You are searching the wrong column. The scope here is one column by design. That is what stops a country fix from rewriting company names, but it also means a value living two columns over is genuinely out of range. Check the column selector before assuming the text is absent.

Whitespace is common enough to deserve its own pass rather than a chain of replacements — the guide to stripping spaces in Excel covers which ones TRIM catches and which need a targeted swap.

Why Excel can find a value it refuses to replace

There is an asymmetry in Excel’s dialog that catches people out and has nothing to do with their search text. The Look in box offers Formulas, Values, Notes, or Comments on the Find tab — but Microsoft’s documentation states that “only Formulas are available on the Replace tab”. Find can look at what is on screen; Replace can only look at what is underneath it.

That produces the confusing case where a search locates a value and the replace then reports nothing to change. A cell displaying 12.50 may store 12.5, and a cell reading 03/04/2026 stores a date serial, not those characters. Find with Look in: Values matches the display; the moment you move to the Replace tab that option is gone and the search runs against the stored value instead. The text you can see was never the text being searched.

The same rule bites from the other direction. Because Replace always reads formulas, a replace-all for a short token rewrites it inside every formula that contains it — useful when you are repointing a model on purpose, damaging when you meant to correct a label and quietly changed cell references across the sheet.

A CSV has neither layer. There are no formulas and no display formats, so the characters in the file are the characters the search compares, and what the preview shows is what the downloaded file contains. If a column arrives as dates or trimmed numbers, that happened when a spreadsheet opened the file — opening a CSV in Excel without mangling it covers the import route that avoids it.

Find and replace in Google Sheets — and when regex is the answer

Google Sheets sits between Excel and this tool. Its find-and-replace dialog offers “Match case”, “Also search within formulas”, and full regular-expression search, which makes it the better choice when the thing you are replacing is a pattern rather than a value: every four-digit year, any phone number in brackets, every trailing punctuation mark. Excel gives you two wildcards; Sheets gives you the whole regex grammar.

The cost is the round trip. Using Sheets means the file goes to Google Drive first, and for a column of customer names, emails, and phone numbers that is a copy on infrastructure you do not control — which is exactly the transfer a browser tool avoids, because the CSV is parsed inside the tab and never sent anywhere. So the honest split is: reach for Sheets when you truly need a pattern, and use this when the text is known and the data is personal. A middle case worth naming — repeated values that differ only by punctuation or case — is often better fixed by normalizing the casing than by writing an expression, and deduplicating afterwards usually collapses what is left.