The number alone tells you very little
Open the report and look at what matched rather than at the total. A long reference list, correctly quoted material, standard methodological phrasing and your own previously submitted work all register as similarity while being entirely legitimate, and together they routinely account for most of a score.
The reverse also holds. A low overall figure can sit beside a genuine problem, because a paraphrase following a source's structure sentence by sentence may not match textually at all while still failing to be your own account of what that source said.
Patchwriting is the real fault, and it is fixable
The common problem in student work is not copying but patchwriting: keeping a source's sentence structure and substituting synonyms. It reads awkwardly, it frequently slips past the tool, and experienced markers recognise it immediately because the register shifts whenever the source is in view.
The remedy is to work from notes rather than from the open source. Read the passage, close it, write what it said in your own construction, then reopen it to check you have it right and add the citation. That produces genuine paraphrase and it is considerably faster than hunting for synonyms.
None of this is about avoiding detection
This page explains what the report measures and how to correct genuine attribution problems. It does not discuss evading review, and nobody in this hall will. Where a match is real, the fix is attribution: quote it and cite it, or rewrite it properly from notes.
Where an institution asks about the use of language models, the answer is disclosure in whatever form they specify. Concealment moves the entire risk onto the person whose name is on the work while the provider offering it keeps the fee, which is the whole of that transaction.
Three steps, then the record keeps itself.
Send the report and the draft
The full report rather than the percentage, and the paper it came from. The matches are what carry the information.
See what actually matched
Each match classified: quotation, reference list, standard phrasing, or something that genuinely needs correcting.
Fix the real ones properly
Quoted and cited, or rewritten from notes so that the account in your paper is genuinely your own.
Queries on record.
Is there a score I should be aiming for?
No, and treating a number as a target produces exactly the wrong behaviour. Programs differ in what they consider worth investigating and many look at matches rather than totals. Twenty percent that is all quotation and references is fine; eight percent containing one unattributed paraphrase is not, and no threshold captures that.
My reference list is matching. Is that a problem?
Almost never, and instructors are thoroughly used to seeing it. Many systems can be configured to exclude the bibliography from the calculation. Where a report is dominated by reference-list matches, the number worth looking at is what remains once those are set aside, which is usually much lower and far more informative.
What separates paraphrase from patchwriting?
Paraphrase restates a source's meaning in your own construction; patchwriting keeps the construction and changes the words. The practical test is whether you could have produced that sentence without the source open in front of you. If you needed it there to get the word order, it is patchwriting and it needs redoing from notes.
Can you help me bring a score down?
Not as an objective, because framing it that way produces the wrong work entirely. What is offered is reading the report to separate legitimate matches from real attribution problems, then fixing the real ones by quoting and citing or rewriting from notes. The number moves as a consequence of that rather than as the goal.
What about AI-detection flags?
The same principle applies: provenance, and disclosure where your institution asks for it. Detection tools produce false positives and no honest service claims to defeat them. What genuinely helps is being able to show your process — drafts, notes, sources you can talk about — which is a by-product of working carefully rather than something constructed afterwards.
Elsewhere in the hall.
Send the assignment and the deadline.
A specialist credentialed in your own field reads what you actually have and replies within two hours with a plan and a price. The reading costs nothing, and if the honest answer is that we are not the right hall for it, you will get that instead.
