What is the Difference Between Turnitin Similarity Score and AI Detection Score?
Table of Contents
- Direct Answer
- How Does Turnitin Calculate the Similarity Score?
- What Types of AI Writing Does Turnitin's AI Detection Identify?
- How Can Checking Both Scores Before Submission Help You Avoid Surprises?
- FAQ
- Sources
- Related articles
Direct Answer
Turnitin provides two distinct metrics in its originality and integrity reports: the Similarity Score and the AI Detection Score. The Similarity Score measures the percentage of a student's text that matches content in Turnitin's vast databases—including web pages, previously submitted student papers, and academic publications. The AI Detection Score, by contrast, measures the percentage of qualifying prose text that Turnitin's model predicts was generated by an AI writing tool, such as ChatGPT, Claude, or Gemini. These two scores are independent of each other: a document can have a high Similarity Score and zero AI detection, or a high AI Detection Score and zero similarity matches [1]. Understanding what each metric captures—and what it does not—is essential for students who want to interpret their reports accurately before submission.
How Does Turnitin Calculate the Similarity Score?
The Similarity Score—often expressed as a percentage from 0% to 100%—represents the proportion of text in a submitted document that matches sources within Turnitin's proprietary databases. These databases include billions of current and archived web pages, millions of previously submitted student papers from institutions worldwide, and tens of thousands of major academic publications, journals, and periodicals [2]. When a document is submitted, Turnitin breaks the text into small chunks and compares each chunk against these databases using advanced fingerprinting algorithms. The resulting percentage reflects how much of the document text overlaps with existing sources.
The score is displayed with a color-coded indicator: blue (0% matching), green (1–24%), yellow (25–49%), orange (50–74%), and red (75–100%). These color bands help instructors quickly gauge the level of textual overlap. Notably, instructors can customize the report by excluding quotes, bibliographic references, and small matches (e.g., fewer than a certain number of words), which can significantly lower the displayed percentage [2]. This flexibility is important because a high Similarity Score does not automatically indicate plagiarism—it may simply reflect proper use of common phrases, direct quotations with citations, or a standardized methodology section.
The Similarity Score does not evaluate whether text was generated by AI, nor does it judge intent. It is purely a text-matching tool. A student who writes an entirely original essay from scratch will typically receive a low Similarity Score (often 0–15%), provided they do not incorporate extensive quoted material. Conversely, a paper that uses AI generation but contains no copied text from existing sources could receive a 0% Similarity Score while having a high AI Detection Score [1][2].
What Types of AI Writing Does Turnitin's AI Detection Identify?
Turnitin's AI detection capabilities have evolved significantly since their launch. The AI detection model is trained to identify text generated by a wide range of large language models (LLMs), including GPT-3, GPT-3.5, GPT-4, GPT-4o, GPT-5 and its variants, ChatGPT (both free and Plus tiers), Claude Sonnet, Gemini (Pro and Flash variants), and LLaMA-based tools [1]. The model processes a submission by breaking it into overlapping segments of approximately a few hundred words (roughly five to ten sentences) and scoring each segment between 0 (human-written) and 1 (AI-generated). The overall percentage reflects the average of all segment scores across the qualifying prose in the document [1].
Beyond standard AI-generated text, Turnitin's current detection also identifies two subcategories. AI-generated only text (highlighted in cyan) is text that the model determines was likely produced directly by an LLM with minimal or no modification. AI-generated text that was AI-paraphrased (highlighted in purple) is text that was likely AI-generated and then further modified by an AI paraphrasing tool such as Quillbot or an AI bypasser/humanizer tool [2]. This two-category breakdown gives instructors a more granular view of how AI tools were potentially used in the writing process.
A critical detail for students is that the AI Detection Score applies only to qualifying text—prose sentences in a long-form writing format (essays, dissertations, articles). Bullet points, code, poetry, scripts, annotated bibliographies, and text under 300 words are not assessed [2]. Additionally, any AI detection percentage between 0% and 20% is displayed as an asterisk (*%) rather than as a specific number. This is by design: Turnitin's testing found higher rates of false positives below 20%, so the asterisk serves as a caution that the score in that range is less reliable and should not be used as sole grounds for academic action [1][2]. The AI Detection Score is completely independent of the Similarity Score—neither metric influences the other [2].
How Can Checking Both Scores Before Submission Help You Avoid Surprises?
Running a pre-submission check on your own document—before it reaches your instructor's Turnitin account—gives you a clear picture of how both the Similarity Score and the AI Detection Score will appear on the official report. This is valuable because the two metrics can tell very different stories about the same paper. For example, you might have a low Similarity Score (indicating original writing with proper citations) but a high AI Detection Score if portions of your draft were composed with the assistance of an AI writing tool [1]. Conversely, you might have heavy quoted material driving up your Similarity Score while the AI score remains at 0%.
Knowing both numbers in advance allows you to take corrective action before the submission deadline. If the Similarity Score is higher than expected, you can identify which passages are matching existing sources and decide whether to paraphrase, cite more thoroughly, or restructure the content. If the AI Detection Score shows flagged text, you can review the highlighted passages to understand what the model is detecting and decide whether to revise the language to reflect your own original phrasing and analysis [4]. Checking pre-submission also prevents the stressful scenario of discovering an unexpected score only after your instructor has already accessed the report.
Turnitin's own guidance emphasizes that the AI indicator should not be used as the sole basis for academic decisions; rather, it serves as a data point for informed discussion between instructors and students [1]. By checking both scores ahead of time, you equip yourself with the same data your instructor will see. This puts you in a stronger position to proactively address any concerns, ask clarifying questions, or make revisions—turning what could be an unpleasant surprise into a manageable, informed part of your writing process.
Many students use turnitin0.com to preview both their Similarity Score and AI Detection Score before submitting to their institution. By uploading your document to turnitin0.com, you receive the same Turnitin similarity and AI writing reports that your instructor would see—complete with color-coded breakdowns, highlighted matches, and the AI detection percentage. This lets you review exactly where citations may need improvement or why the AI detector is flagging certain passages, all before the final submission. No subscription is required, and your document is never archived or sent to any third-party database.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
FAQ
Does a high Similarity Score automatically mean plagiarism?
No. A high Similarity Score indicates that text in your document matches content in Turnitin's databases, but it does not determine intent. Properly cited quotes, commonly used phrases, and correctly formatted bibliographies can all generate matches. Instructors review the highlighted matches and use their professional judgment to distinguish between proper citation practices and academic misconduct [2].
Can the AI Detection Score flag text I wrote myself?
Turnitin reports a false positive rate of less than 1% for its AI detection model. However, the model is less reliable for scores below 20%, which is why those scores are displayed as an asterisk (*%) rather than a specific number [1]. If you believe your text has been incorrectly flagged, you should discuss the report with your instructor, who can review the highlighted passages in context.
Is the Similarity Score affected by AI detection, or vice versa?
No. The two scores are completely independent. A document can have a 0% Similarity Score and a 100% AI Detection Score if it is entirely AI-generated but contains no text matching existing sources. Conversely, a heavily quoted paper could have a high Similarity Score and a 0% AI Detection Score [2]. Neither metric influences the calculation of the other.
What file types does Turnitin accept for AI detection?
Turnitin processes files in.docx,.pdf,.txt, and.rtf formats for AI writing detection. The submission must contain at least 300 words of prose text in a long-form writing format (essay, dissertation, article) and must not exceed 30,000 words or 100 MB. Bullet points, code, poetry, and tables are not processed as qualifying text [2].
Can students check their own Similarity and AI scores before submitting to an instructor?
Turnitin's institutional reports are typically visible only to instructors and administrators through the learning management system. However, some services allow students to submit their documents and receive the same Turnitin-generated reports privately, including both the similarity analysis and the AI writing indicator, before the official submission [1][4].
Sources
- Turnitin's AI Writing Detection Capabilities FAQs — https://guides.turnitin.com/hc/en-us/articles/28477544839821-Turnitin-s-AI-writing-detection-capabilities-FAQs
- Using the AI Writing Report — https://guides.turnitin.com/hc/en-us/articles/22774058814093-Using-the-AI-Writing-Report
- How AI Detection Works — https://helpcenter.turnitin.com/hc/en-us/articles/27811948436237-How-AI-detection-works
- Discussing AI Writing with Students — https://www.turnitin.com/blog/academic-integrity-and-ai-writing-how-to-discuss-with-students
Related articles
- Canada Essay Deadlines: Running Similarity + AI Checks the Night Before (Safely)
- How Do I Reduce the Ai Score on My Dissertation Without Changing My Argument?
- Turnitin False Ai Flag
- Can Turnitin Really Detect Chatgpt? What the AI Report Actually Measures
- Humanizing Tools and Academic Tone: Preserving Claims, Evidence, and Hedging Language