Turnitin Similarity Colors, Explained

Turnitin colors your similarity score — and the color carries meaning before you read a single number. The five-color scale runs blue to red; here's the full chart, what each band means for essays and papers, and the color most students misunderstand.

The Turnitin similarity color chart

ColorSimilarity rangeWhat it means
BlueNo matching text (0%)No significant overlap found — including quoted material in some configurations
Green1-24%The normal zone for most student work; usually no questions
Yellow25-49%Worth a look — check what matched before anyone asks you
Orange50-74%Heavy overlap; expect scrutiny and prepare explanations
Red75-100%Near-total overlap; serious conversation ahead

Ranges reflect Turnitin's standard color bands; individual instructors can interpret them differently. Note this is the similarity (text-overlap) indicator — a different number entirely from the AI writing score.

Is green safe? Is yellow bad?

Green (1-24%): for a typical essay, yes — comfortable territory. A green score made mostly of properly cited quotations and bibliography entries is routine. Blue (0%) surprises students: zero matching can simply mean short text, obscure sources, or quotes excluded by settings — it is not automatically suspicious, and it is not automatically clean of AI either (the two scores are independent).

Yellow (25-49%): the decisive factor is the shape of the matches. Spread across twenty cited sources: usually defensible. One 25% block from a single source: that's the pattern that draws questions. See what a good score actually is and the honest fix-it path in is 25% bad?

Before you submit: shrink the avoidable matches

Most yellow scores come from patchwriting — swapping words into source sentences. Paraphrase by meaning instead, quote sparingly, cite as you write. If matched passages trace to AI-drafted text, that's the other score: run the AI check and fix flagged lines before submitting. Both numbers, one free tool.

Know your score before they do

Free, unlimited AI detection with per-line flags — no login, no submission repository, nothing stored.

Check your text free →