Texto y contenido

Zipf's Law Text Analyzer

Paste a text to get a rank–frequency chart on log-log axes, the slope, R² and a Zipf fit index, plus the words that appear noticeably more often than Zipf's law predicts.

Text to analyze

Remove stop words (ru, uk, en, pl)

0 / 50 000 caracteres

Everything is computed in your browser — the text is not sent to our server.

How the numbers are calculated

Words: runs of Cyrillic and Latin letters (hyphens and apostrophes inside are kept). Numbers and punctuation are dropped, case is ignored, Russian «ё» is treated as «е».
Slope and R²: a linear regression on log(rank) vs log(frequency). Only words seen at least twice are used (or the top 50 words if fewer than 10 qualify): the tail of one-off words forms a flat “step” that would distort the slope.
Index 0–100 = 60% × how close the slope is to −1 (1 at −1, 0 when off by 0.8 or more) + 40% × line fit (1 at R² = 1, 0 at R² ≤ 0.7). 80+ is a close fit, 60–79 moderate, below 60 weak.
The “Zipf” column is the classic f₁ / rank, where f₁ is the count of the most frequent word; “Deviation” is the difference from it. The “above norm” mark is relative to your text’s own regression line instead, so a shallow or steep slope alone doesn’t turn every word into an outlier: the word is not a stop word, occurs at least 3 times and 1.5× more often than the line predicts. A rank-1 word is also marked when stop words are kept and it still outranks all of them: in natural text the top spots belong to articles, prepositions and conjunctions.

Datos clave

Zipf’s law: in natural text a word’s frequency is roughly inversely proportional to its rank, and on a log-log chart this is almost a straight line with a slope near −1. Paste a text and the tool draws the chart, computes slope and R² and gives a fit index. It is statistics, not an AI detector: a short text, a list or mixed languages will deviate.

Precio: Gratis
Registro: No necesario
Procesamiento: En tu navegador: no se envía nada a nuestro servidor
Entrada: Text of 50 words up to 50,000 characters
Límites: No sign-up and no request limits; at least 50 words and at most 50,000 characters; estimates are unstable on texts under 300 words
Resultado: The slope and R² of the log-log line, a Zipf fit index from 0 to 100, a top-30 word table with frequencies and stop-word marks, a rank–frequency chart with the ideal line and a list of words more frequent than predicted
Última actualización: 2026-10-12

Zipf's Law Text Analysis Online

Check a text against Zipf's law: the tool counts word frequencies, plots rank vs frequency on log-log axes next to the ideal Zipf line, computes the slope and R² of a linear regression and combines them into a 0–100 fit index. Words that appear noticeably more often than the text's own distribution predicts are highlighted, which helps spot keyword stuffing. Runs in your browser, no sign-up.

log(rank) vs log(frequency) chart with the ideal Zipf line and a regression line

Slope, R² and a 0–100 Zipf fit index with a published formula

Top 30 words: count, share, Zipf-expected count and deviation in %

Highlights words used more often than expected

Total and unique words, lexical diversity (TTR)

Optional stop-word removal for English, Russian, Ukrainian and Polish

Cómo usar

Paste a text — 300+ words, ideally 1,000+ (up to 50,000 characters)
Optionally turn on stop-word removal and click “Analyze”
Review the slope, R², the chart and the highlighted words in the table

FAQ

An empirical pattern: if you sort words by frequency, the frequency of the nth word is roughly inversely proportional to its rank. The most frequent word occurs about twice as often as the second and three times as often as the third.

On a log-log rank–frequency chart natural text lies almost on a straight line with a slope near −1. The closer your slope is to −1 and the higher the R², the closer the distribution is to the classic one.

No. Fitting Zipf’s law is statistics, not a detector: a high index does not prove human authorship and a low one does not prove machine authorship. Short and specialised texts deviate.

Prepositions and conjunctions fill the top of the table and hide meaningful words. Without them you see which content words repeat most. Stop words are marked in the table.

At least 50 words, but estimates are stable only over hundreds and thousands of words. For texts under 300 words the page warns that the slope is unstable.

PromoPilot frente a alternativas: comparación

Comparamos solo lo que está confirmado en las páginas de cada servicio. Donde no pudimos confirmar un dato, la celda muestra "—".

Parámetro	PromoPilot	Wikipedia: закон Ципфа (справка)
Rank–frequency chart with slope and R²	✓	Explains that a log-log plot shows how closely data follow the law; no tool
Fit index 0–100	✓	✗
Words more frequent than expected	✓	✗
No text sent to a server	✓	—

Desplaza la tabla hacia los lados

Dónde somos más débiles: tokenisation is simple with no lemmatisation, up to 50,000 characters, no file upload or multi-text comparison, and the index is our own scale, not a standard metric.

Cuándo elegir PromoPilot: When you want to see the word-frequency distribution of your own text in seconds, spot overused words and judge how natural the text looks. A single index is not enough to judge AI authorship or SEO text quality.

Comprobado el: 2026-10-16. "—" significa que no está confirmado en la página oficial; las condiciones y los precios cambian, así que consulta los sitios de cada servicio.