Set up one clear replacement
Paste your source into Text, enter the exact passage in Find, and enter the new passage in Replace with. Choose whether to match capitalization and whether matches must be whole words, then select Replace all. The result shows the transformed text and a replacement count. Review both before copying. This operation changes every accepted match in the supplied text, not just the first one you notice.
Use a specific search when the source contains several contexts. Replacing a short fragment can affect longer words, identifiers or addresses that happen to contain it. Searching for a full phrase is often more controlled than searching for one common word. For example, a project label with its surrounding punctuation is more distinctive than a two-letter abbreviation that also appears inside ordinary sentences.
The original field remains available after the operation. The result is a separate read-only output, so running again still uses the input rather than automatically using the previous result. To perform a second transformation on the output, copy it back into the input deliberately. Keep a separate original when several changes must be reviewed together; this page does not provide a saved revision history.
Changing the text, search, replacement or options clears the earlier result. Press Replace all again to produce a result for the new settings. This avoids copying an output that belongs to a previous search. Clear removes the input, the search and the replacement text. Before processing another document, also inspect the two matching options so they suit the new task.
Literal search means punctuation is ordinary text
The Find field searches for a literal sequence. A dot means a dot, an asterisk means an asterisk, and brackets mean brackets. They are not commands for matching arbitrary characters. This makes the page useful for replacing punctuation, template markers and small fragments of plain text without learning a pattern language. It also means a regular expression copied from a programming tutorial will not behave as a regular expression here.
Try the source “a.b a?b a.b”, search for “a.b”, and replace it with “$&$1$$”. With the initial matching options, the result is “$&$1$$ a?b $&$1$$” and the count is two. The question-mark version stays untouched because it is not the same literal sequence. The example also shows that dollar signs and apparent replacement tokens remain exactly as typed.
In some editors, a replacement such as $& inserts the matched text, or $1 refers to a captured group. That is not how this tool works. Here those strings are ordinary replacement content. You can insert a price marker, a dollar sign or a template token without it being expanded into another value. There are no capture groups because there is no regular-expression search mode.
The same caution applies to backslash notation. Typing a backslash followed by n is not a request to insert a new line through a special escape command. The search and replacement controls are single-line fields. For formatting the beginning and end of separate lines, use the line-formatting tool instead of trying to construct a pattern with anchors or escape sequences.
Decide whether capitalization must match
Match case is off initially. In that state, a search can match uppercase and lowercase forms under the browser's Unicode case-insensitive matching rules. Turning Match case on requires the capitalization to match the search. This option changes which passages qualify; it does not change the capitalization of the replacement to imitate the original passage.
For example, use “Test test TEST” as the source, “test” as the search and “note” as the replacement. With Match case off, all three occurrences become note. With Match case on, only the middle occurrence changes, giving “Test note TEST”. Choose the second behavior when an uppercase abbreviation or a title-case name must remain distinct from a lowercase word.
If you want replacements that preserve sentence capitalization, plan separate passes with Match case on. You might change a capitalized term first and a lowercase term second, inspecting the result between operations. The tool does not infer sentence boundaries or automatically produce Note, note and NOTE from one replacement. A single replacement value is inserted identically at every accepted match.
Ignoring case does not mean ignoring accents, correcting spelling or finding synonyms. A search for an unaccented word is not a request to remove diacritics from the source. Visually similar characters and alternative Unicode representations can also differ. When an expected occurrence is missed, copy the intended source fragment into Find and inspect spaces and accents before making the search broader.
Whole words use boundaries, not a dictionary
Whole words only is also off initially. Without it, a match may occur inside a longer sequence. With it on, the tool checks the characters immediately before and after each possible match. Letters, numbers, combining marks and underscores count as word characters. A match touching one of those characters is skipped. The rule works with Unicode letters rather than treating every accented character as a separator.
Consider the source “café caféine décafé café!” and search for “café”, replacing it with “tea”. With Whole words only enabled, the result is “tea caféine décafé tea!” and the count is two. The standalone word and the word before an exclamation mark change. The occurrence joined to more letters on either side does not qualify as a whole-word match.
Punctuation such as a hyphen can create a boundary, while an underscore or digit does not. This matters for identifiers and compound labels: a search might match one part of a hyphenated phrase but skip the same letters in an underscore-separated code. The option is a character-boundary check, not a grammar engine that understands which compound terms should remain intact.
For languages that commonly write several words without spaces, this boundary rule may not match your linguistic expectations. You can turn the option off and use a longer exact phrase to control the replacement. Do not use whole-word matching as proof that only one concept was changed. Read each affected context, especially when a term is part of a name, an address or a technical identifier.
Remove content and interpret the count
Delete a matching passage
Leaving Replace with empty deletes every accepted match. Find must still contain something; an empty search produces an error rather than matching between every character. This is useful for removing a known marker from a draft. Include any surrounding spaces in the search only when you intend to remove those spaces too, and check that the remaining words do not run together.
For instance, use “Test test TEST”, search for “test”, turn Match case on and leave the replacement empty. The result is “Test TEST”, with two spaces between the remaining words. The count is one. The tool removed the matching letters, not the neighboring spaces. It does not automatically tidy spacing, rewrite the sentence or remove the line where the match appeared.
Read the replacement count
The count reports accepted occurrences, not the number of words changed or the size of the visible difference. A phrase containing several words can count as one replacement. If the replacement is identical to the matched text, the operation can count a match even though the output looks unchanged. Zero means no accepted occurrences were found under the selected options, and the text remains as supplied.
Matches do not overlap. Searching for “ana” in “banana” and replacing it with “x” produces “bxna”, with one replacement, not two overlapping replacements. Newly inserted text is not searched again within the same operation either. Replacing a with aa does not continue expanding forever. These rules make one pass predictable, but the order of several separate passes can still change the final result.
Review a batch edit before using it elsewhere
Before a broad replacement, test a short sample containing a normal occurrence, an occurrence inside a longer word and a capitalization variant. Those examples reveal whether the two options match your intention. Then run the full excerpt and compare the count with what you expected. An unexpectedly large count is a reason to inspect the search, not to assume the tool found extra corrections you had missed.
For a useful review, compare the original input with the output using Compare texts. This makes unintended changes easier to locate. Read the complete sentences around the edited terms: changing a noun can require a different article, verb or plural ending that a literal replacement will not supply. The tool performs the requested substitution, not an editorial rewrite.
When multiple replacements depend on each other, choose their order carefully. Changing a long phrase first can preserve context that would disappear if a short word were replaced first. Copy each accepted result into the next input deliberately and keep track of the settings. There is no multi-rule queue or automatic mapping table on this page, so each pass contains one search and one replacement.
Use Copy result after checking the output, then paste into the intended destination and inspect it there. Plain text does not preserve a source document's comments, styles or table structure. Replacing text that happens to resemble a URL, formula or command does not validate that structure. Do not treat a successful replacement as proof that the resulting link, expression or code fragment will work.
Know the limits and choose another task when needed
The source text is limited to 20,000 units according to the field counter. The search and replacement values are also checked against a 20,000-unit limit, and output is guarded at 200,000 units. A short source with many matches can produce a large result when the replacement is long. If the output limit is reached, shorten the replacement or process a smaller corresponding section.
These limits use the browser's string length. Some emoji and combining sequences occupy more than one unit, so the count can differ from the number of visible symbols. A failed operation does not provide a complete transformed document. Adjust the input and run again, then inspect the new result before copying. Do not combine an earlier successful output with a failed later attempt as though they belonged to one pass.
If the task is to add the same label before every line or wrap each item with punctuation, use Add prefix and suffix. It works on line boundaries without needing to search for existing text. This page is for replacing content that is already present. It does not sort lists, remove duplicate entries, correct spelling or provide a regular-expression editor.