Color bands on the similarity report train people to panic at yellow. The number is coverage of matching strings, not a moral score. A methods paragraph copied from the lab handout can outrank a stolen essay that was heavily paraphrased.
How is the percentage counted?
Turnitin concatenates matching spans and divides by document length. Exclude bibliography and exclude quotes if your instructor enabled those filters. If they did not, your perfectly cited block quote still paints the bar.
| Score | Common story | What to check first |
|---|---|---|
| 0 to 10% | Short paper or heavy original wording | Still open matches. A stolen abstract can hide in a small %. |
| 11 to 24% | Quotes, prompts, and shared templates | Quoted vs unquoted. Template language from the assignment. |
| 25 to 40% | Needs a human look | Long unmatched-to-you blocks. Missing citations. |
| 40%+ | Often a problem, or a take-home with a huge prompt pasted in | Did you paste the questions? Did you submit the wrong file? |
Does submitting to class add your paper to the database?
Usually yes for a standard LMS assignment. That is how later classes get caught recycling a roommate’s file. It is also why a “practice submit” on a random website can haunt the real due date: your draft becomes a source.
A private, non-repository scan is a different product. Hero HW Services sells that kind of check so you can see overlap without depositing the draft. Your school’s official assignment is still the one that archives.
People also ask
- Is 0% similarity the goal?
- Almost never. Real papers cite sources. Zero can look as suspicious as 70% if the assignment required literature.
- Does similarity detect ChatGPT?
- Not by itself. ChatGPT text that was never published anywhere can score low on similarity and still trip the separate AI writing indicator.