What is Turnitin similarity reduction, and what is it not?
It is reading the report, sorting every match into the kind of overlap it actually is, attributing the ones that should be attributed, and rewriting only the ones that were genuinely unattributed. It is not a service for pushing a number down, and we will say so rather than take the brief. The reason is that a similarity score is a total and a total is not a diagnosis: one figure covers required boilerplate, properly marked quotation, your own earlier work, technical artefacts like reference lists and tables, and genuine unattributed borrowing, and those need opposite responses. A service that only lowers the number will paraphrase your own faithful quotation, hide your own prior work, and leave a genuine borrowing intact because it happened to sit under the threshold.
What similarity percentage is safe to submit at?
There is no stable answer, and anyone who gives you a number is guessing. Departments set different thresholds, some set none at all and decide from the report rather than the figure, and the number changes with the field: a technical document with formulae and a standard instrument will always score higher than a reflective essay on the same subject, without either being at fault. It also depends on the corpus and the version, which change without notice, so a target quoted this term may not be reproducible next term. That is a far more useful target than a percentage, and it is the one worth working towards.
Why is my score high when I did not plagiarise?
In practice, because most of the total in a well-written document is legitimate and cannot be removed without damaging the work. A named instrument or scale everyone uses. A sentence of theory you must quote to state the theory faithfully. All of these produce matches, and all of them should, because the point of a definition or an instrument is that everyone words it identically. Add to that a technical artefact, since reference lists and appendices are counted in some configurations, and a document that is a model of careful attribution can still report a figure that looks alarming. It is also worth checking whether the report excludes the sections you assume it does.
How do I reduce my similarity score legitimately?
By adding attribution rather than removing words, and by dealing only with the matches that are yours to fix. Restore the quotation marks to any passage that was a direct quotation and lost them, and add the citation, which lowers your responsibility while leaving the wording in place. Cite the source of a definition, of a reproduced table, of an instrument, and of a scale, and where a document is being analysed rather than used, say so in a sentence. Then deal with the genuine unattributed borrowing by quoting it, paraphrasing it properly, or cutting it. Three of those steps add no new words at all, which is why blanket rewording is both slower and worse.
What is self-plagiarism, and is it a problem?
It is reusing your own earlier material in a new document without saying so, and the usual sources are a published article, a thesis or project submitted elsewhere, a paper written for another module, or a conference version. It is a real concern in most fields, because the reader of the new document has no way to know the material appeared before, and because a large body of work presented as newly written is misleading even when every word is yours. It is not fixed by rewording, and rewording is its own kind of dishonesty, in the opposite direction: it conceals a fact that is better stated.
Does quoting with quotation marks still count as similarity?
A quoted passage will register as a match whatever you do to it, and the quotation marks and the citation are what convert that match from a problem into evidence that you handled the source properly. The alternative is worse in every direction: rewording a quotation so it stops reading as one converts documented borrowing into unmarked borrowing, which is a genuine misconduct problem rather than a cosmetic one. The related distinction is that a paraphrase should change the structure and the sentences, not merely the order of the words, because light reordering leaves enough of the original visible that it is easy to read as unattributed copying.
Does the reference list count towards my similarity score?
In many settings the reference list, the bibliography and the appendices are excluded from the calculation, and in some they are not, which means a document whose sources happen to be the sources everyone else cites can report a noticeably higher figure for no fault at all. The vendor's own published material sets out what is excluded in the version your institution uses, and that is worth reading directly rather than assuming. The broader point is the one this page keeps making: a reference list matching other documents is evidence that you read the same literature, not a defect.
Can I be wrongly accused of plagiarism by a similarity report?
Yes, and it is not a theoretical risk. A report measures textual overlap against a corpus, and it has no access to intent, to whether a passage was marked as a quotation, or to whether the overlap was required by the field, so a high figure can accompany a document in which nothing improper happened. The flip side is equally important: a low figure does not establish that a document is clean, because a paraphrase written closely enough can avoid the comparison altogether. If you are being questioned, the useful steps are to produce your drafts in order, produce your own dated record of prior work, and ask for the passages actually identified rather than arguing with a total.
Will you spin or rephrase my work to lower the number?
No, and this is the clearest thing we decline. Spinning, swapping or light rephrasing in order to move a figure is the practice this page argues against, and it does harm: required definitions become less precise, faithful quotations stop reading as quotations, and your own prior work stops looking like yours. We will also not reword another author's ideas until they stop registering as a match, since that is not paraphrasing and it is the version most often requested under that name. What we do instead is find the matches that are genuinely unattributed and deal with those, and leave your required wording, your citations and your own earlier work alone.
Do you help with the whole report, or just the number?
The whole report, which is the only version of this that is worth having. A number cannot be worked on directly, because it is an average over categories that need different treatment, and the categories are visible in the source list and the highlighted passages. So the work is: every source in the report resolved to something you can read, every match classified as required, marked quotation, your own earlier work, technical artefact, or genuine unattributed borrowing, and then an action appropriate to each. You get the classification document as well as the revised text, which means you can see the reasoning and argue with it, and which is what makes the work useful to you rather than merely reassuring.
What if my university has a specific threshold?
Then the threshold tells you how much time to spend rather than what to do, and it is worth establishing exactly what it means before you begin. Some departments treat a percentage as a trigger for conversation; some require a report review; some set a figure above which work is checked by a second reader; and some change the threshold between cohorts, so the number you were told last year may not be the one applied now. Find out which of these applies to you, in writing if you can, because a departmental policy and a rumour about a tool lead to very different experiences.
Where is Writely available, and how do I start?
Writely is based in Nigeria and works remotely with students, researchers and candidates across Nigeria and the rest of Africa, including learners who need similarity classification and attribution support. Use the
Chat with us button to send the document and the full report, not just the percentage, along with any earlier work of yours that may overlap and the threshold or rule your department applies. Saying whether you used a named instrument or scale is genuinely useful, because that single fact explains a large share of the total in most technical documents. Also tell us the deadline, and the team can respond with a suitable next step, including a plain statement of what we will and will not take on.