Skip to content

How to Check Text Similarity: Plagiarism Detection Guide

What Is Text Similarity?

Text similarity is a numerical measurement of how much two documents overlap in content. Unlike a diff checker that shows exact changes between two versions of the same document, a similarity checker compares two potentially unrelated texts and produces a percentage score — 0% means completely different, 100% means identical.

Check the similarity between any two texts with WritePadPro's Text Similarity & Plagiarism Checker — paste both documents and get a similarity score, with the matching phrases highlighted in both texts. All processing runs locally in your browser.

Text similarity has applications across many fields: academic integrity (detecting student plagiarism), SEO (identifying duplicate content across web pages), legal review (comparing contract clauses), journalism (verifying original reporting), and content marketing (ensuring your content is distinct from competitors). The core question is always the same: how much do these two texts share?

How N-Gram Jaccard Similarity Works

WritePadPro uses n-gram analysis combined with the Jaccard similarity coefficient — a proven, well-understood algorithm for measuring text overlap.

Step 1: Generate N-Grams

An n-gram is a contiguous sequence of n words from a text. For a 3-gram (trigram) analysis of the sentence "The quick brown fox jumps over the lazy dog," the n-grams are:

  • "the quick brown"
  • "quick brown fox"
  • "brown fox jumps"
  • "fox jumps over"
  • "jumps over the"
  • "over the lazy"
  • "the lazy dog"

WritePadPro lets you choose 2-, 3-, 4- or 5-word phrases (3 is the default), and it ignores capitalization and punctuation. Each n-gram captures a small phrase of context. Using n-grams instead of individual words is critical because two texts about the same topic may share many individual words ("the," "is," "writing") without actually being similar in content. Shared n-grams indicate shared phrasing — a much stronger signal of genuine overlap.

Step 2: Build N-Gram Sets

The algorithm generates the complete set of n-grams for both texts. Text A might produce 500 unique trigrams. Text B might produce 480 unique trigrams. The algorithm then identifies the intersection — n-grams that appear in both sets.

Step 3: Calculate Jaccard Coefficient

The Jaccard similarity coefficient is:

J(A, B) = |A ∩ B| ÷ |A ∪ B|

Where |A ∩ B| is the number of n-grams shared by both texts, and |A ∪ B| is the total number of unique n-grams across both texts.

If Text A has 500 trigrams, Text B has 480, and they share 120 trigrams, the Jaccard similarity is: 120 ÷ (500 + 480 − 120) = 120 ÷ 860 = 0.1395 = 13.95%

Why Jaccard Works Well

The Jaccard coefficient is bounded between 0 and 1 and is symmetric (comparing A to B gives the same result as B to A). It has one important limit: when one text is much shorter than the other, the score stays low even if the short text was copied in full, because the union is dominated by the longer text. That is why WritePadPro also shows how much of each text matched, as a percentage of the words in Text A and in Text B. It does not require a database of web content — it compares exactly the two texts you provide, making it well suited for local, privacy-preserving analysis.

Text Similarity vs Plagiarism Detection

Text similarity and plagiarism detection are related but not identical concepts. Understanding the distinction prevents misinterpretation of results.

Text Similarity Is a Measurement

A similarity score is a neutral number. Two texts sharing 40% of their n-grams is a fact, not a judgment. The texts might be similar because one copied from the other (plagiarism), because both drew from the same source (common sourcing), because they cover the same narrow topic (topical similarity), or because they use standard phrases from a shared field (domain-specific language).

Plagiarism Is a Judgment

Plagiarism is the act of presenting someone else's work as your own without proper attribution. It requires intent and context that no algorithm can determine. A similarity checker provides evidence that a human can use to investigate, but the tool itself does not declare plagiarism. A high score between a student essay and a Wikipedia article deserves a close look. A high score between two contracts built from the same standard boilerplate can be expected and innocent.

What WritePadPro's Tool Does and Does Not Do

WritePadPro's Text Similarity & Plagiarism Checker compares two specific texts that you provide. It does not search the internet or check against a database of published content. This makes it ideal for:

  • Comparing a student's essay against a specific source you suspect was used
  • Checking if two articles on your site are too similar (duplicate content)
  • Verifying that a rewritten article is sufficiently different from the original
  • Comparing a paraphrase against its original to see how much phrasing carried over

For internet-wide plagiarism detection (checking if content exists elsewhere on the web), you would need a service with a web crawl database. WritePadPro serves the local comparison use case — two documents you have, compared privately.

Interpreting Similarity Percentages

The similarity score requires interpretation based on context. WritePadPro labels each score with one of five bands.

WritePadPro's Score Bands

Similarity ScoreLabelWhat It Usually Means
Under 10%Minimal SimilarityThe texts are largely independent.
10-25%Low SimilaritySome common phrases, often coincidental.
25-50%Moderate SimilarityNoticeable overlap; the texts may share source material.
50-80%Significant SimilaritySubstantial shared content.
80% and aboveVery High SimilarityNearly identical, or one text is derived from the other.

Context Matters More Than the Number

There are no standard thresholds for "acceptable" similarity. The same pair of texts scores higher with 2-word phrases than with 5-word phrases, and a short passage copied into a long document can still score low. Two news reports on the same event share names and facts; contracts built from the same template share boilerplate; an original essay should share little phrasing with any single source. Check the highlighted phrases and the matched-words percentages before drawing conclusions.

For exact word-by-word differences between specific document versions, use the Text Diff Checker instead — it shows precisely what changed. For more on comparing versions, see our Document Comparison Guide.

Common Use Cases

Academic Integrity

Educators paste a student submission alongside a suspected source text. If the score is higher than you would expect for original work, look at the highlighted phrases. The tool helps instructors provide specific evidence when discussing academic integrity with students — pointing to exact shared phrases rather than making vague accusations.

SEO Duplicate Content Detection

Google has said that duplicate content on a site is not grounds for action unless the intent is to deceive or manipulate search results, but near-duplicate pages compete with each other, and Google may show only one of them. Compare pages on your site that target similar keywords to make sure they are distinct. If two blog posts share a lot of phrasing, consider consolidating them into one comprehensive piece or differentiating their keyword targets. For condensing overlapping content, see our Text Summarization Guide.

Content Rewriting Verification

After rewriting a piece of content (for syndication, localization, or improvement), check the similarity between the original and rewritten versions. If long runs of the original phrasing are still highlighted, the rewrite may be too superficial — more paraphrasing or structural reorganization is needed.

Translation Quality Assessment

Compare a back-translation (translated text re-translated to the original language) against the original. Shared phrasing shows where wording survived the round trip. A low score does not prove meaning was lost, because a faithful translation can use different words, so read the differences yourself.

Ghostwriting and Attribution

Compare a suspect text against an author's known body of work. If an allegedly original article shows high similarity to previously published content by someone else, it raises attribution questions. Combined with stylometric analysis (word frequency patterns), similarity checking can add evidence in authorship disputes. For more on catching common writing issues, see the Grammar Mistakes Guide.

The Privacy Advantage of Local Processing

Online plagiarism checkers that search the web have to send your text to their servers for analysis. This creates serious privacy and intellectual property concerns:

  • Unpublished work exposure — your draft novel, research paper, or business plan is sent to a third-party server
  • Data retention — some services store submitted texts in their databases (to compare against future submissions)
  • Legal documents — uploading confidential contracts to a plagiarism service may violate NDA terms
  • Student data — some services have faced criticism for building proprietary databases from student submissions

WritePadPro's Text Similarity & Plagiarism Checker runs the entire n-gram Jaccard analysis locally in your browser using JavaScript. Your text is never transmitted anywhere. The computation happens on your device, and the data disappears when you close the tab. This makes it safe for:

  • Confidential legal documents and contracts
  • Unpublished manuscripts and research
  • Business-sensitive content
  • Student work (no contribution to third-party databases)
  • Any text you would not want stored on someone else's server

Using WritePadPro's Text Similarity Checker

WritePadPro's Text Similarity & Plagiarism Checker provides instant, private text comparison.

Step 1: Open the Tool

Navigate to the Text Similarity & Plagiarism Checker. Two text panels are shown: Text A (Source) and Text B (Compare).

Step 2: Paste Both Texts

Enter the first text in panel A and the second text in panel B. These can be any length — from a paragraph to a full article. The order does not matter (Jaccard similarity is symmetric).

Step 3: Run the Analysis

Choose a phrase size (2 to 5 words; 3 is the default), then click Check Similarity. The tool builds the n-gram sets for both texts and calculates the Jaccard coefficient. You see the score and its label, the number of unique phrases in each text, the shared phrases, how much of each text matched, and the matching phrases highlighted in both texts. Copy Report copies a plain-text summary.

Step 4: Interpret the Score

Use the score bands above, then read the result in context: the same score can be harmless for two posts on the same topic and worth a closer look for a student essay compared with a source.

Step 5: Investigate If Needed

If the score raises concerns, read the highlighted phrases in both panels to see where the overlap is. When the two texts are versions of the same document, the Text Diff Checker shows exactly what was added, removed or kept.

Summary

Text similarity analysis measures the degree of overlap between two documents using n-gram Jaccard similarity — a proven algorithm that compares shared word sequences to produce a percentage score. The score ranges from 0% (completely different) to 100% (identical).

Interpreting the score requires context: the same overlap can be normal for two blog posts on the same topic and worth investigating between a student essay and a published source. The tool measures similarity — it does not determine plagiarism, which requires human judgment about intent and attribution.

Check text similarity with WritePadPro's Text Similarity & Plagiarism Checker — n-gram Jaccard analysis running 100% in your browser. Your text is never transmitted to any server, making it safe for confidential, unpublished, and legally sensitive documents.

Frequently Asked Questions

How is text similarity different from plagiarism?

Text similarity is a neutral mathematical measurement — the percentage of shared content between two documents. Plagiarism is a human judgment about intent — presenting someone else's work as your own without attribution. A similarity score of 40% between two texts is a fact. Whether that 40% constitutes plagiarism depends on context: it could be common topic coverage, shared boilerplate language, legitimate quotation with citation, or actual unauthorized copying. The similarity checker provides evidence; a human determines whether that evidence indicates plagiarism.

What similarity percentage indicates plagiarism?

There is no universal threshold. The score depends on context, on the phrase size you choose and on how long each text is: a short passage copied into a long document can still score low, which is why WritePadPro also shows how much of each text matched. News reports on the same event share facts and names, contracts built from the same template share boilerplate, and an original essay should share little phrasing with any one source. Always read the highlighted phrases — the number alone does not determine plagiarism.

Does WritePadPro check against the entire internet?

No. WritePadPro's Text Similarity Checker compares two specific texts that you provide — it does not search the internet or check against a database of published content. This design is intentional: it ensures complete privacy because your text never leaves your browser. For internet-wide plagiarism detection, you would need a service with a web crawl database (like Turnitin or Copyscape). WritePadPro serves the local comparison use case — when you have two specific documents and want to know how similar they are.

What are n-grams and why do they matter for similarity checking?

N-grams are contiguous sequences of n words from a text. A 3-gram (trigram) of 'the quick brown fox' produces: 'the quick brown,' 'quick brown fox.' N-grams matter because comparing individual words produces misleading results — two texts about different topics may have many individual words in common simply because common English words (the, is, and, to) appear in everything. N-grams capture phrases and phrasing patterns, which are much stronger indicators of genuine content overlap. Shared trigrams show shared phrasing, not just shared vocabulary, although stock phrases such as 'the study found' can match by chance — another reason to read the highlighted phrases.

Is my text safe when using WritePadPro's similarity checker?

Yes. WritePadPro's Text Similarity Checker runs the entire n-gram Jaccard analysis locally in your browser using JavaScript. Neither text is ever transmitted to any server, stored in any database, or accessible to anyone other than you. The computation happens entirely on your device. When you close the browser tab, all data is gone. This makes it safe for unpublished manuscripts, confidential contracts, student submissions, proprietary business content, and any text you would not want on a third-party server.

Try These Tools

Related Articles