What is the Difference Between the Turnitin Similarity Score and the AI Score?
Table of Contents
- How Do Turnitin Similarity Score and AI Score Define Different Types of Academic Integrity Issues?
- What Do the Percentages in a Turnitin Similarity Report Versus an AI Writing Report Actually Measure?
- Should I Check Both My Similarity Score and AI Score Before Submitting an Assignment?
- FAQ
- Sources
- Related articles
Direct Answer - The Turnitin Similarity Score and the Turnitin AI Score measure two separate dimensions of academic integrity. The Similarity Score checks your submitted text against Turnitin's databases of web pages, academic journals, and previously submitted student papers to identify matching or unoriginal content. The AI Score, by contrast, analyzes whether portions of your text were likely generated by an artificial intelligence tool such as ChatGPT, Claude, Gemini, or other large language models [1]. These two scores operate completely independently — a paper with a high Similarity Score can have a low AI Score (for example, properly cited direct quotations from published sources), while a paper with a low Similarity Score can have a high AI Score (original-sounding text that was entirely AI-generated). Understanding this distinction is essential for any student who wants to submit work with confidence.
How Do Turnitin Similarity Score and AI Score Define Different Types of Academic Integrity Issues?
The Similarity Score and the AI Score each address a fundamentally different category of academic conduct. The Similarity Score is rooted in source attribution — it identifies text that matches content already existing in Turnitin's databases, including web content, academic publications, and other student submissions [2]. A high similarity percentage suggests that the writer may have used sources without proper citation or paraphrasing, which falls under traditional plagiarism concerns.
The AI Score, in contrast, targets authorship authenticity. It examines the statistical patterns of word choice and sentence structure to determine whether text was likely produced by an AI writing tool rather than a human writer [1]. Where the Similarity Score asks "did you copy from somewhere else?", the AI Score asks "did a machine write this for you?" These are distinct questions that require different responses from both students and instructors.
Because the two scores measure independent issues, a submission can present challenges in one area without any concern in the other. A student who carefully cites every source but uses AI to draft their analysis may receive a low Similarity Score alongside a high AI Score. Conversely, a student who writes entirely from their own thinking but fails to cite paraphrased sources may receive a high Similarity Score with a low AI Score [2]. Recognizing which type of integrity concern applies to your own work is the first step toward addressing it effectively.
What Do the Percentages in a Turnitin Similarity Report Versus an AI Writing Report Actually Measure?
The percentage displayed in the Similarity Report represents the proportion of your submission's text that matches content in Turnitin's databases. A 25% Similarity Score, for instance, means that one-quarter of the document contains text that matches existing sources — this could include properly quoted passages, poorly paraphrased sections, or directly copied material [3]. The Similarity Report highlights each matched segment and links to the original source, giving both students and instructors a clear picture of where and how sources have been used.
The AI writing indicator percentage, by contrast, represents the proportion of the document that Turnitin's detection model predicts was generated by an AI writing tool. The model works by breaking the submission into segments of roughly a few hundred words (approximately five to ten sentences), scoring each segment between 0 (human-written) and 1 (AI-generated), and averaging those scores across the entire document [1]. A 40% AI Score means that the model predicts approximately 40% of the document's prose was produced by an AI system.
It is critical to understand that these two percentages are computed through entirely different methodologies. The Similarity Score relies on a direct database comparison — it looks for literal textual matches. The AI Score relies on a statistical model trained to detect differences in word probability patterns between human and AI writing [3]. Human writing tends to be inconsistent and idiosyncratic, producing low-probability word sequences, while AI-generated text tends to follow highly probable, consistent patterns. These fundamentally different detection methods mean that the two scores cannot be compared, combined, or substituted for one another.
Should I Check Both My Similarity Score and AI Score Before Submitting an Assignment?
Yes — checking both scores before submission is strongly recommended because they cover distinct and independent aspects of academic integrity. Relying on only one score leaves a significant blind spot. If you check only the Similarity Score, you may submit work that is entirely original in terms of source use but was generated by AI, potentially violating your institution's AI use policy. If you check only the AI Score, you may submit work that appears human-written but contains unattributed source material, risking a plagiarism finding [4].
The independent nature of these two scores means that each requires its own review strategy. For the Similarity Score, you should examine highlighted matches to confirm that all sources are properly cited and paraphrased. For the AI Score, you should review flagged segments to understand why they were identified as potentially AI-generated, especially if you used AI tools as part of your writing process [4]. Some institutions permit limited AI use with disclosure, while others prohibit it entirely — knowing your AI Score allows you to make informed decisions about your submission.
Self-checking both scores also gives you the opportunity to address issues before your instructor sees them. Turnitin's Similarity Report and AI writing report provide the same data that instructors will review, so a pre-submission check allows you to see exactly what they will see. By understanding both the percentage and the highlighted content in each report, you can take corrective action — whether that means revising citations, rewriting flagged passages, or disclosing permitted AI use — before your work is officially submitted [1][4].
Understanding how your Similarity Score and AI Score work together is only the first step. The next step is seeing your own actual Turnitin reports before your instructor does — so you can review your similarity matches, check your AI writing indicator, and make any needed revisions with confidence. Turnitin0 gives you access to the same institutional-grade Similarity Report and AI writing report that your university uses.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
FAQ
Can a paper have a high Similarity Score but a low AI Score?
Yes. A paper containing properly attributed direct quotations from published sources may show a high similarity percentage while scoring very low on AI detection, since the text was written by a human and only the quoted portions match existing sources.
Can a paper have a low Similarity Score but a high AI Score?
Absolutely. If a student uses AI to generate entirely original-sounding text that does not match any existing database content, the Similarity Score may be very low, but the AI Score may be high because the writing patterns are consistent with AI generation [1].
Are the Similarity Score and AI Score calculated the same way?
No. The Similarity Score uses direct text matching against Turnitin's content databases. The AI Score uses a statistical model trained on word probability patterns to distinguish human writing from AI-generated text [3]. The methodologies are entirely separate.
Do instructors see both scores together?
Yes. Turnitin displays the AI writing indicator alongside the Similarity Report, allowing instructors to view both scores in the same interface and evaluate a submission's integrity from both angles [1][2].
Should international or second-language writers be concerned about false positives?
Turnitin trained its AI detection model on a representative sample that includes second-language writers and diverse geographic and subject-area data to minimize bias [1]. While no detection system is perfect, the model is designed to reduce misidentification for these groups. Turnitin reports a false positive rate of less than 1% for its AI detection indicator.
Sources
- Turnitin's AI Writing Detection Capabilities FAQs — https://guides.turnitin.com/hc/en-us/articles/28477544839821-Turnitin-s-AI-writing-detection-capabilities-FAQs
- AI Writing Detection and the Similarity Score: What Educators Should Know — https://www.turnitin.com/blog/ai-writing-detection-and-the-similarity-score-what-educators-should-know
- AI Writing Detection: Facilitating Honest Conversations with Students — https://www.turnitin.com/blog/ai-writing-detection-facilitating-honest-conversations-with-students
- Academic Integrity Best Practices for Students Using AI — https://www.turnitin.com/blog/academic-integrity-best-practices-for-students-using-ai