Text Statistics Analyzer Tool

1. 1

   Paste or open your text

   Type or paste directly, or open a .txt, .md, .csv, .json, .srt or .vtt file. Nothing is uploaded - the analysis runs on your device.

2. 2

   Read the counts and the reading level

   Characters, words, sentences, paragraphs, syllables and pages update as you type, alongside six readability formulas and a consensus grade.

3. 3

   Check the words and phrases you repeat

   The frequency table shows each term's count and its density as a percentage. Switch phrase length to two, three or four words to see repeated wording rather than repeated words.

4. 4

   Look a term up in context

   Select any word or phrase in the table to see every place it occurs with the text either side of it, so you can decide whether the repetition is a problem.

5. 5

   Export the report

   Download TXT for a readable summary, CSV to open the tables in a spreadsheet, or JSON to keep the numbers and the settings that produced them.

The analyzer reports thirty-one numbers about a piece of text, in four groups. Every one of them is computed on your device as you type; nothing is sent anywhere, and there is no queue and no watermark.

Counts

• Characters, three ways: the raw length, the length without spaces, and the number of visible characters. Those last two differ more often than people expect - an emoji is two units of raw length and one visible character, and so is an accented letter typed as a base plus a combining mark.
• Words, sentences, paragraphs and lines: a paragraph is a block of lines with a blank line between it and the next; a line is a line break.
• Syllables and pages: syllables feed four of the reading-level formulas, and pages are counted at 250 words each, the double-spaced convention every submission guideline uses.
• Unique words and vocabulary richness: how many distinct words the text uses, and that number divided by the total. A long text with a low ratio is repeating itself.

Time estimates

• Reading time defaults to 200 words per minute and is adjustable, with a faster-to-slower range so you can plan for an audience rather than an average.
• Speaking time defaults to 130 words per minute, which is a measured presentation pace rather than a conversational one.
• Handwriting time uses 20 words per minute, the middle of the range measured for continuous longhand.

Frequency and phrases

• The word table shows each term's count and its density as a percentage of the whole text, which is the number every SEO tool quotes and the one a bare count cannot give you.
• The phrase table shows repeated two-, three- or four-word runs. Phrases never cross a sentence boundary, so nothing in the table is an accident of two sentences sitting next to each other.
• The character table counts every letter and digit in any script, not just a to z.

Readability formulas turn sentence length and word length into a school grade. They disagree with each other on purpose - each was built for a different kind of writing - so the analyzer reports all six and takes the median of the five grade scores as a consensus. Below about ten words no formula means anything, and the tool says so rather than printing a number.

Flesch Reading Ease

A score from 0 to 100 where higher is easier. Around 60 to 70 is standard English prose; below 30 is graduate-level. It is the only one of the six that is not a grade, which is why it is not averaged into the consensus.

Flesch-Kincaid Grade Level

The same inputs as Reading Ease, expressed as a US school grade. The most widely quoted grade score, and the one most word processors report.

Gunning Fog

Adds a penalty for words of three or more syllables. It runs high on technical writing, which is the point - it was designed to catch business prose that had drifted into jargon.

SMOG

Counts words of three or more syllables and normalises to thirty sentences. It is the standard in health and medical writing, where the cost of a reader not understanding is highest.

Coleman-Liau

Uses letters per word rather than syllables, so it needs no syllable estimate at all. That makes it the most reliable of the six on text where syllable counting is shaky.

Automated Readability Index

Also letter-based, and was built to be computed by a typewriter-era machine. It tends to run a grade above Coleman-Liau on the same text.

Readability checker

Most online counters use two regular expressions and hope. The differences are small on a tidy English paragraph and large on anything else, so here is exactly what this one does.

Words

A word starts with a letter or a digit and continues with letters, digits and combining marks. That last part matters: combining marks are not letters, so a pattern that only looks for letters cuts Thai, Devanagari, Arabic with vowel points and decomposed accents mid-word. Chinese, Japanese, Thai, Khmer, Lao and Burmese have no spaces between words, so those runs are segmented with your device's own Unicode word-breaking rules rather than treated as one long word. You can decide whether numbers count, and whether contractions and hyphenated compounds stay whole.

Sentences

Splitting on . ! ? counts "It costs 1,000.50 dollars." as three sentences and "The U.S.A. is big." as four. The default here uses the Unicode sentence-breaking rules, which get decimals, web addresses and initialisms right on their own, plus a list of abbreviations per language for the cases the rules cannot know - "Dr." is never the end of a sentence, while "etc." sometimes is. The old punctuation behaviour is still available if you need a number that matches another tool.

Characters

The raw character count is the number of storage units, which is what a database field limit counts. The visible-character count groups combining marks, emoji sequences and flags into the single characters a reader sees. A thumbs-up with a skin tone is four storage units and one visible character. Both are shown because both are the right answer to a different question.

Word counter · Text statistics

Density is a term's count divided by the total word count. There is no target number - the idea that some percentage is optimal has not been true for a long time - but a term at four or five per cent in a page of prose is usually a sentence that could be rewritten. The phrase table is more useful than the word table for this, because writing repeats phrases long before it repeats words.

• Turn on "ignore common words" to drop articles, pronouns and prepositions. The list holds roughly 150 to 190 terms per language, not a dozen, so the table actually changes.
• A phrase is dropped if it starts or ends with a common word, so "of the" never reaches the table while "peace of mind" does.
• Phrases that occur once are not shown. One occurrence is a sentence fragment, not a phrase.

Word frequency counter · Keyword extractor

A frequency table tells you a word appears forty times. It does not tell you whether those forty are spread through the piece or crammed into one paragraph, and that is the difference between a theme and a tic. Selecting any word or phrase opens a concordance - every occurrence with the text either side of it, in document order. This is a standard tool in corpus linguistics and it is the fastest way to decide whether repetition is a problem worth fixing.

The analyzer groups every sentence into a length band and shows the distribution, alongside the longest and shortest sentence in the text. Prose that reads well moves between bands; prose where every sentence lands in the same band reads flat however good the individual sentences are. An average cannot show you this - a text of all twenty-word sentences and a text alternating five and thirty-five have the same average and read nothing alike.

The tool is built for the moment before you publish, submit or send something.

• Meeting a word limit: essays, applications, abstracts and product copy, with the count updating as you cut.
• Checking a draft is readable: six grade scores and a consensus, so you are not trusting a single formula's quirk.
• Auditing content for repetition: density and phrases together show what a page actually keeps saying.
• Timing a talk or a script: speaking time at a real presentation pace, not a reading pace.
• Sizing subtitles: open a .srt or .vtt directly and get the word count and reading level of what viewers actually see.
• Comparing several drafts: analyze up to twenty-five documents and export one CSV row per document to sort in a spreadsheet.

Every number here is either exact or estimated, and the difference is worth knowing before you quote one.

Exact

Characters, words, sentences, paragraphs, lines, unique words, frequencies and densities are counted, not estimated. Two independent implementations of the whole analysis are checked against each other on every release across tens of thousands of comparisons, and they have to agree exactly.

Estimated

Syllables cannot be counted exactly without a pronunciation dictionary, so they are estimated by rule. Measured against a hand-counted list of 145 words the estimator is exact on 144 of them, and its total across the list is within half a per cent. The one it misses is a twenty-eight-letter joke word. The four formulas that use syllables inherit that error, which is well inside the width of a grade band.

What it does not do

• It does not check grammar or spelling, and it does not rewrite anything.
• The reading-level formulas are calibrated on English. They produce a number for any language and that number means progressively less the further you get from English.
• It reads plain text. Word documents, PDFs, slides and images need converting first - the links below do that on this site, also on your device.

One megabyte of text is analysed in well under a quarter of a second, off the main thread, so typing never stutters.

The analyzer reads files whose bytes are the text: .txt .text .md .markdown .log .csv .tsv .json .xml .html .htm .srt .vtt .rtf. Anything else is refused at the file picker rather than accepted and then failed, and the refusal links to the tool on this site that turns it into text.

For everything else:

• Word to text
• PDF to text
• Extract text from PowerPoint
• Excel to CSV
• Image to text
• Speech to text

Every measurement on this page is free. There is no metric, no formula, no phrase table and no export that Pro unlocks - the plan difference is scale, which is where the cost actually is.

• Free: one document up to 10 MB, every statistic, all six formulas, the concordance, and TXT, CSV and JSON export. The top twenty-five words, phrases and characters - more than the ten or fifteen the best-known competitors show.
• Pro: up to twenty-five documents at once with one combined CSV and all the reports in a single ZIP, the complete frequency table rather than the top twenty-five, and no size cap beyond what your device can handle.

• Word counter
• Readability checker
• Word frequency counter
• Keyword extractor
• Text statistics
• Case converter