How to Read a Turnitin Similarity + AI Report — What Every Number Means (2026)

Updated 2026–2027 | United Kingdom

You submitted your assignment. Now you're staring at a Turnitin report with a percentage in the corner, maybe an asterisk next to something labelled "AI," and absolutely no idea whether you should be worried.

I spent eleven years supervising dissertations. I've sat across from students who panicked over a 38% similarity score that was completely fine — and students who were confident about a 12% score that had a serious problem buried inside it. The number alone tells you almost nothing. What matters is knowing how to read it.

This guide explains every number, colour, and symbol in your Turnitin report — and what it actually means for your submission.

Part 1

The Similarity Score

What It Is — and What It Isn't

The similarity score is the percentage of your submission that Turnitin's algorithm matched against its databases: previously submitted student papers, academic journals, websites, and published books.

The formula is simple: matched words ÷ total words = similarity percentage.

What it is not is a plagiarism score. Turnitin itself is clear on this point — it detects similarity, not intent. A 40% score might be entirely made up of correctly formatted citations. A 10% score might contain a stolen paragraph from a blog post Turnitin didn't happen to index.

The Colour Codes Explained

What's a "Normal" Score in Your Discipline?

There is no universal threshold. What's normal varies significantly by field:

What matters more than the percentage is where the matches appear. A 40% score concentrated entirely in your reference list is a non-issue. A 15% score with a matched paragraph in your argument section is worth a closer look.

What Turnitin Does and Doesn't Index

Turnitin's database is large but not infinite. It includes:

It does not reliably index: books not digitised in its database, some paywalled journal content, content in languages other than English (though coverage is improving), and most internal university documents.

Part 2

The AI Detection Score

How Turnitin's AI Detection Actually Works

Turnitin's AI writing detection analyses your prose for two primary signals:

The AI score represents the percentage of your "qualifying text" (prose sentences above a minimum length threshold) that Turnitin predicts was generated by an AI tool.

Score Ranges and What They Mean

The Asterisk: Why Scores Below 20% Are Hidden

Since July 2024 — and reaffirmed in Turnitin's August 2025 policy update — any AI detection score between 1% and 19% is displayed as an asterisk (*%) rather than a number.

The reason is false positives. Scores below 20% carry a meaningfully higher risk of incorrectly flagging human-written text, particularly for non-native English writers and highly technical or formulaic writing styles. Turnitin made this change specifically to avoid potential false positive incidents at low confidence thresholds.

If you see an asterisk, here is what it means in practice:

The University of Melbourne's official guidance for educators confirms this: an asterisk indicates the score was below the 20% display threshold, and no text will be highlighted within the report.

False Positives — They Are Real

Turnitin reports a false positive rate of under 1% for documents where at least 20% of the content is AI-generated. However, independent research and educator reports suggest this rate can climb in specific conditions:

If you are wrongly flagged: gather your drafts, version history, and research notes. Request a meeting with your instructor. Most university policies — and Turnitin's own guidance — are explicit that an AI detection report alone is not sufficient evidence of academic misconduct. A second piece of evidence is required before any formal process begins.

Part 3

The AI Writing Report — Highlights and Breakdown

Segment-Level Analysis

The AI Writing Report doesn't just give you a single score. It analyses each sentence individually, assigning a confidence score between 0 (very likely human-written) and 1 (very likely AI-generated). Sentences above a confidence threshold are highlighted in the full report.

This matters because it lets your instructor (and you, if your university grants access) identify exactly which parts triggered the detection — not just that something was flagged overall.

Two Colours, Two Different Problems

The August 2025 update introduced a second detection category, bringing the total to two:

Part 4

Reading Both Reports Together

The similarity score and AI detection score measure completely different things. They are independent — a paper can be entirely original in the plagiarism sense (no matched sources) while showing high AI detection, or it can be full of correctly cited quotes (high similarity) with no AI signals at all.

Common Scenarios

What Your Instructor Actually Does With These Numbers

Most instructors use the report as a starting point for a conversation, not a judgment. Standard practice at most UK and Australian universities (where Turnitin is most widely used) is to:

Part 5

Exclusions — Why Your Score Might Be Lower Than Expected

Turnitin allows instructors to apply filters that remove certain content from the similarity calculation:

If these exclusions are applied, your displayed similarity score will be lower than the raw figure. This is by design — it strips out correctly formatted citations and common short phrases that would otherwise inflate the percentage artificially.

If you're unsure whether exclusions are applied to your submission, check with your instructor or look at your university's Turnitin configuration settings.

Part 6

The Problem Students Don't Know About

Here is something the official documentation doesn't make obvious: students typically cannot see their own AI Writing Report.

Turnitin does not make the AI detection score or highlighted report visible to students by default. You can submit an assignment, receive a similarity score, and have no idea your AI report flagged 60% of your submission — until your instructor contacts you.

Some universities allow students to view similarity scores before final submission (usually by enabling multiple submission attempts). Almost none grant access to the AI Writing Report on the student side.

This means you are going in blind.

The practical solution is to run your own check before your university does.

Premier Check — £2.99 Free AI Content Detector

Frequently Asked Questions

The Bottom Line

Turnitin reports are data, not verdicts. A similarity score tells you where your text overlaps with existing sources — not whether you plagiarised. An AI score tells you how your writing patterns compare to known AI outputs — not whether you used AI.

Once you understand what each element actually measures, the report stops being intimidating and becomes exactly what it was designed to be: a tool to help you improve your work.

If you want to know what your instructor will see before they see it — run your own check first.

Last reviewed: June 2026 · Turnitin's interface and policies are subject to change. Always verify current settings with your institution.