A Secondary 3 student in Clementi submits an English assignment and sees a large similarity percentage beside the file. The essay was written at the dining table over three evenings. There are notes, a messy opening paragraph, and a second draft in which the student changed the order of the argument. Yet the report paints entire lines in bright colours. Is the report saying the student cheated? Is the work no longer their own? And should a parent insist that the child rewrite every highlighted sentence until the number falls?
A plagiarism checker, similarity checker, or originality report can be useful in school, but the three labels are not interchangeable. The software can find passages that resemble material in its comparison collection. It cannot, from a percentage alone, decide why that resemblance exists, whether a quotation was attributed correctly, whether an idea has been misappropriated, or whether the school’s particular rules have been broken. A match is an observation. Misconduct is a judgement that requires context. That difference changes how students should write, how teachers should review reports, and how families should respond when a score looks alarming.
Here is the interesting part: a genuinely original essay can contain matched sentences, and a heavily copied argument can produce a deceptively modest similarity score. The result is not a paradox once we separate text overlap from intellectual ownership. Let us open the report, trace the mechanisms, and make a decision that would stand up to a calm conversation at school.
The 50-second answer
A similarity system compares submitted writing with sources it can access, including webpages, publications, and—in some products or institutional settings—previously submitted papers. It flags overlapping text and often summarises how much of the submission is matched. A large match may come from a copied paragraph, but it may also come from correctly quoted evidence, a shared assignment question, a standard definition, a reference list, or the student’s own earlier submission. A small score is not proof of originality: the student might have copied an idea without copying its wording, or the relevant source might not be in the database. Read the actual matched passages, inspect the sources, check quotation and attribution, and apply the school’s published assessment rules. Never decide guilt or innocence by a traffic-light colour alone.
Quick route: If you are a student, open the largest individual matches and identify what each passage is doing. If you are a parent, ask to see the marked text and the assignment requirements, not just the percentage. If you are a teacher, separate screening evidence from a fair academic-integrity finding and record the reason for any conclusion.
First, define the question the tool actually answered
Imagine an essay about whether schools should provide additional green spaces. The writer has been taught to introduce a claim, give a reason, consider a counterargument, and conclude. The whole class is using the same assignment question and the same two supplied readings. Several students use a sentence such as “Green spaces can support both recreation and biodiversity.” That sentence may appear on many sites, especially when the topic is discussed in similar educational language. None of this automatically demonstrates dishonesty. The sentence is general enough that independent writers could construct it; it could also be copied. The report does not tell us which explanation is correct.
The system has answered a narrower question: where does this text resemble text already available for comparison? Depending on the platform and settings, the matching process may identify exact strings and more flexible kinds of textual overlap. The report then associates spans of the student’s text with sources. Its number is a summary of matched material under a particular configuration. It is not a percentage probability that the student cheated. It is not an ethical score. It is not an examiner’s mark for quality.
Turnitin’s published guidance states explicitly that its Similarity Report compares writing against a database and highlights matching text rather than determining plagiarism. This is an important provider-specific description, not a guarantee that every commercial checker has the same coverage, exclusions, or algorithms. Some products market themselves more strongly than their measurement can justify. A responsible school asks what the product actually detects, how the matching rules work at a high level, what student work is retained, and whether staff can inspect the original comparison passages.
We should also distinguish plagiarism from an academic integrity violation more broadly. Plagiarism ordinarily concerns presenting another person’s words or ideas as one’s own without appropriate acknowledgement, but institutional definitions differ. Unauthorised collaboration, prohibited AI assistance, fabrication of sources, or submitting someone else’s work may breach rules even when a similarity engine finds very little matching language. Conversely, quoting a poem with accurate citation may be entirely appropriate and may generate a conspicuous block of matching text. The software is one source of evidence about the document, not the rulebook for the entire assignment.
Mechanism one: matching words is easier than determining who owns an idea
Why do matching tools focus on words? A computer can compare sequences of symbols with large indexed collections at speed. It can locate passages with overlapping wording, identify the apparent sources, and let a reviewer move through the document. Detecting intellectual dependence is a different task. Two writers can express the same borrowed argument with very different words; two independent writers can express an ordinary fact with nearly identical words. The path from strings to intentions and authorship is therefore not one-to-one.
Consider three invented lines. Student A writes, “The library was busier after school because the weather was wet.” Student B writes, “Rainy afternoons sent more pupils indoors, and the library filled up.” Student C copies the first sentence word for word from an online article but removes its attribution. A and B may be describing the same observation independently, while C has reused another author’s expression without acknowledgement. Yet a purely verbal overlap detector might regard A and C as more alike, even though their writing histories differ. That is precisely what the detector is built to find: similarity of text, not direct access to the history of creation.
The distinction becomes clearer with a scientific idea. “Photosynthesis converts light energy into chemical energy” is a familiar educational explanation. A student’s unremarkable use of that established concept is not the same thing as claiming to have discovered it. But if the student copies a researcher’s distinctive analysis, sequence of evidence, or novel argument and merely substitutes synonyms, there can be an attribution problem even if the similarity percentage stays low. Ethical writing must preserve the provenance of borrowed ideas as well as the wording of quotations.
This creates two kinds of error for readers. A false suspicion occurs when a legitimate match is interpreted as proof of wrongdoing. A false reassurance occurs when a small percentage is interpreted as proof that no problem exists. Teachers and students need habits that reduce both errors. The useful follow-up question is never simply “How can I get a lower number?” It is “Can I explain where each substantive idea came from, what language is mine, and what the assignment allowed?”
Mechanism two: the percentage is a summary with hidden choices
Suppose an invented 1,000-word assignment produces 240 words of identified matches. The displayed score might be 24%, depending on how that system counts overlapping segments, excluded sections, and the submission’s text. This arithmetic is illustrative, not a disclosure of any product’s proprietary method. Change the composition of the same assignment and the interpretation changes immediately. If 160 matched words are the prescribed essay question and assessment rubric pasted into the first page, much of that percentage may carry little information about the student’s authorship. If 160 matched words are a complete uncited analysis lifted from an essay site, the concern is different.
Even identical percentages can conceal different patterns. One paper may have twenty tiny matches of standard phrases spread across the introduction, headings, and references. Another may reproduce a continuous block of distinctive argument. A third may include a correctly cited 200-word primary-source quotation that the task explicitly requires students to analyse. All three could display similar headline numbers and demand different responses. The colour scale is a navigation aid. It does not establish a shared ethical meaning for every file in the same band.
Small matches also depend on thresholds. A product may allow a teacher to exclude quotations, bibliographies, or sources below a specified length. Excluding appropriate material can make review more focused, but changing these filters changes the displayed score without changing a single word of the essay. Teachers should therefore avoid comparing raw percentages collected under different settings as though they were measurements made with identical instruments. A report should be read with its date, scope, comparison sources, and filter choices intact.
The comparison database matters too. A document cannot match a private source that the engine has no permission or technical means to examine. A newly posted webpage may not yet be indexed. A student’s previous paper may be stored in one institutional repository but not another. Some tools can report self-matches against earlier submissions. None of these conditions means the student changed their behaviour. They mean the observation environment changed. The score is partly a property of the document and partly a property of the system that examined it.
Five common matches that do not settle the case
A supplied question. Some teachers ask students to keep the full prompt at the top of the document. If everyone copies the question into the response, the same wording appears across dozens of submissions. It may inflate the score while telling us little about the essay itself. The repair is administrative: identify or exclude the prompt where the tool permits, and judge the original answer separately. Students should not conceal required instructions merely to reduce a number.
A direct quotation. English and History assignments sometimes require careful analysis of a fixed passage. Quoting that passage with an adequate citation is a legitimate scholarly move when the task permits it. The match may be obvious because the whole point of a quotation is to retain the source’s exact words. The reviewer should inspect the quotation marks, attribution, amount quoted, analysis surrounding it, and assignment rules, rather than treating the match as automatic evidence against the writer.
A reference list. Names of authors, book titles, publication dates, journal names, and persistent identifiers naturally recur. Standard referencing styles encourage consistency, so correctly formed references may look like one another. A bibliography match can still deserve inspection if sources are fabricated or copied without being consulted, but the mere recurrence of citation details is unsurprising. Some tools provide reference-list exclusions; these should be applied transparently, not used to hide substantive borrowing elsewhere.
A standard expression. Definitions, subject-specific terms, formula labels, and common connecting phrases can appear in multiple legitimate essays. A Primary 6 Science student and a textbook may both say that an object casts a shadow when light is blocked. There are only so many efficient ways to state a common elementary mechanism. A single stock sentence is weaker evidence than a long sequence of unusual phrasing, identical examples, and the same conceptual order.
A student’s earlier work. A writer who submitted an initial draft to one classroom portal and a final draft to another may see the final version match the draft. This is sometimes described as self-matching. Whether reusing previously assessed work is permitted is a separate question: some assignments require an entirely new response, while others explicitly expect revision of a draft. Ask about the submission trail and the school’s policy before drawing conclusions from the match alone.
The most dangerous mistake: treating a score as a target
There is a temptation to turn a complex integrity question into a simple instruction: keep the similarity score below a particular percentage. Numbers feel objective. Students can plan around them. Administrators can add them to a spreadsheet. Yet this habit can teach exactly the wrong lesson. A young writer may replace a clear, correctly quoted source with a vague paraphrase that removes the citation. The percentage might fall while the integrity of the work becomes worse. Another may replace established terminology with strange synonyms until a scientific explanation is no longer accurate. The checker becomes a game rather than a guide to responsible writing.
A score threshold can sometimes be part of a local administrative screening workflow, but it is not a universal safe harbour. Schools may use rules, provisional flags, or subject-specific instructions that differ. A Literature assignment containing lengthy quotations is unlike an original reflective journal. A science report may use method vocabulary that appears in many lab manuals. A creative-writing submission may contain very little reason for direct quotation. Compare reports only after considering what the task expects and what kinds of legitimate overlap are normal for that genre.
Imagine a teacher telling a class that every essay must be “under 15%.” A student writing a close reading of a poem quotes six short phrases accurately, uses the required textual evidence, and includes a bibliography. The report reaches 17%. Another student avoids quoting the poem, makes broad assertions about its meaning, and reaches 4%. Which student wrote the stronger literary analysis? The answer cannot be read from those two numbers. The teacher needs to look at the validity of claims, use of evidence, attribution, interpretation, and independent thought. An arbitrary percentage rule could inadvertently reward an essay that is less precise.
The right question is whether a tool changes behaviour in ways that strengthen learning. If students understand why borrowed expression needs quotation marks and why borrowed ideas need attribution, the report can make revision concrete. If students learn only to reduce coloured highlights, the tool may push them away from academic honesty. A good classroom policy explains the purpose of screening, how matches are interpreted, what a student can do before submission, and how concerns will be discussed. The policy protects both intellectual work and the learner who is still learning to write.
A worked Clementi example: four essays, one headline percentage
Consider an imaginary Secondary 4 English group at a tuition table in Clementi. The students answer: “Should public spaces prioritise quiet study over lively community activity?” Every example below is fictional; the numbers are teaching devices, not observed student records. The assignment requires one source-based paragraph, a counterargument, and the student’s own conclusion. The teacher receives four reports, each with a 22% similarity headline. Rather than rank the learners by the identical number, she prints the match map and asks what the matched regions are doing.
Essay A: the quotation and the prompt. Twelve percentage points are from the teacher’s supplied prompt and ten from a correctly marked passage in the source packet. The student’s own argument and evaluation are written in original wording. The appropriate next step is to check that the quotation fits the task and that its source is identified. If a school permits exclusions, the score can be recalculated for convenience, but the intellectual judgement does not depend on making the number vanish.
Essay B: the lifted explanation. Most of the matching text is one long paragraph from an online commentary. It follows the source’s unusual examples and distinctive argument, without quotation marks or meaningful citation. This warrants a conversation and closer review. The student may have misunderstood paraphrasing, copied a model answer, or intentionally represented the passage as original. The report raises a question; it does not supply the student’s explanation, intent, or the correct disciplinary procedure.
Essay C: the early draft. Nearly all the flagged passages match an earlier version of the same student’s essay stored in a school repository. The assignment explicitly invited iterative drafting. The teacher can inspect the draft history and revisions. Here, the match is evidence of continuity between versions, exactly what the task requested. The relevant assessment is whether the student made the requested improvements and followed any rules about submitting prior work.
Essay D: the patchwork response. The report shows many brief matches from different sites. The student has stitched together small phrases and sentence fragments, sometimes adding words between them. The total matched percentage is the same as the others, but the writing may show weak synthesis and unclear ownership of ideas. The teacher needs to compare the source passages, ask for notes or a writing explanation if appropriate, and teach the difference between gathering information and constructing an argument. The same score has four different stories.
This exercise can be repeated using unmarked passages that contain no copied text. Students can practise classifying a match as expected, acceptable with proper attribution, needing correction, or needing further investigation. Classification teaches a skill the single score never could: identifying the reason for apparent overlap. It also encourages proportional responses. Some cases need a missing citation repaired; others may need fuller review under school procedures. A teacher’s careful interpretation protects both standards and trust.
What a strong paraphrase really changes
Paraphrasing is not a mechanical substitute operation. The goal is to understand an idea, retain its intended meaning, and express that understanding in a structure chosen for the current argument. For example, an imaginary source might claim that students benefit from quiet library areas because reduced interruptions help them sustain attention during extended reading. A superficial paraphrase swaps “benefit” for “gain” and “interruptions” for “distractions” while preserving the same sentence path. A stronger paraphrase first asks what relationship matters: interruption frequency, attention, and the demands of longer tasks. The writer then chooses relevant terms and connects the source’s reasoning to the essay’s question.
A responsible paraphrase still acknowledges the origin of a borrowed idea. Students sometimes believe that changing every word erases the need to cite. It does not. Attribution serves more than one purpose: credit, traceability, fair representation, and enabling a reader to check the original context. The citation should point to a source the student actually consulted, not an impressive-looking entry copied from somebody else’s reference list. This is particularly important when an AI tool offers a paragraph with citations that the student has not opened. A polished bibliography is not evidence of research unless the listed sources support the claims.
Try this method with a short, permissible source passage. Read it until you can answer, in ordinary speech, what claim the author makes, what evidence is supplied, and what limitation is acknowledged. Close the passage. Write three brief notes: claim, mechanism, boundary. Draft the new paragraph from those notes for the actual assignment question, then reopen the source and compare meaning. If the new wording is too close to distinctive source language, quote or rewrite with genuine restructuring. Cite the idea. This workflow puts comprehension before the checker and makes ethical writing a visible intellectual skill.
A citation is not permission to copy unlimited text. Copyright, assignment instructions, and fair quotation practices still matter, and those are separate from the similarity system’s calculation. Nor does citation automatically make an essay strong. A page of correctly attributed quotations can contain almost no independent analysis. The student should use sources in service of a clear argument, not replace their thinking with a procession of authorities. The goal is not zero overlap; it is accurate dependence, transparent borrowing, and discernible contribution.
How to read a similarity report without panic
Begin with the assignment. Was the task an argumentative essay, a quotation-based analysis, a research summary, a lab report, or an original reflection? What source use did the teacher require? Were collaboration, earlier drafts, or digital assistance permitted? A report without the task specifications is like a temperature reading without the location or instrument. You may have a number, but not enough context to interpret it. Write the task requirements beside the report before treating any colour as meaningful.
Next, inspect where the matches occur. A high total concentrated in the first page’s template is different from a high total inside the student’s analytic paragraphs. Follow each large match to the alleged source. Is the source the teacher’s worksheet, the student’s earlier draft, a public article, or a repository entry? If the original is inaccessible, note that limitation rather than pretending the report proves what the passage contained. If two sources reproduce one another, the checker may name the later copy rather than the earliest author; trace the source chain where authorship matters.
Then inspect the function of each matched passage. Is it a direct quote, a conventional definition, a distinctive explanation, a method, a bibliographic entry, or the student’s original observation? Does the writing put quotation marks around exact words and cite borrowed ideas? Does the source support the claim made in the essay, or has the student taken it out of context? A citation can be formally present yet refer to the wrong page or fail to support the sentence. These are writing questions requiring reading and judgement, not threshold calculations.
Fourth, review the larger composition. Is there an argument the student can defend? Do the examples follow from the evidence? Are the transitions doing original reasoning or merely joining pasted fragments? Can the learner explain a paragraph in their own words and point to notes or revisions? A process record can help resolve uncertainty, but it should not be demanded selectively or treated as a perfect lie detector. Students work differently; accessibility accommodations, multilingual writing, and handwritten planning can leave different kinds of traces. The school should apply fair and consistent standards rather than reward only those who produce a particular electronic history.
Finally, decide a response proportionate to the evidence. A missing citation may be teachable and correctable under classroom rules. A substantial uncited match may need formal review. A score generated largely by the reference list may need no action at all. Record the distinction between a finding, a question needing clarification, and a teaching point. Parents should expect the school to explain its reasoning, not reveal another student’s confidential work. Teachers should expect students to engage with evidence, not insist that every coloured phrase is harmless.
The difference between similarity checking and AI detection
Another source of confusion arrives when families see “AI writing” and “similarity” discussed together. They are different propositions. A similarity checker looks for resemblance between a submitted text and material in comparison collections. An AI-writing detector, where offered, attempts to infer whether a pattern of text was generated by a model. It does not have direct access to a student’s intent or necessarily to the drafting process. Different products have different error characteristics, and neither category should be treated as a universal authorship oracle.
A student who copies a paragraph from a webpage might generate a clear similarity match without having used AI. A student who produces an entirely original argument with authorised assistance from a language tool may have little textual overlap with indexed webpages. A third student might use an AI system to invent a source or to supply an argument they cannot explain, leaving academic-integrity issues that the similarity score alone cannot resolve. Schools need rules that distinguish assistance from substitution, disclose what is allowed, and give students practical ways to demonstrate their learning.
In a Clementi English discussion, a helpful exercise is to take one sentence and ask two questions. First: Where did this language or idea come from? Second: What thinking did the student contribute? Those questions may point toward source acknowledgement, assignment integrity, or both. Neither is reducible to the colour of a report. Teachers can teach ethical use of tools while keeping the evaluation centred on the student’s actual reasoning and the stated learning objectives.
A seven-step student route before submission
Step one: identify the assignment’s evidence policy. Read whether sources are required, whether quotation is permitted, what referencing style is expected, and which kinds of assistance must be declared. Ask early if any rule is ambiguous. Doing so is easier than repairing a whole essay after submission and better than assuming the rules from a previous subject carry over unchanged.
Step two: keep a source trail while researching. Save each useful source’s author or organisation, title, date when available, URL or publication details, and the claim it can actually support. Separate direct quotations from paraphrase notes, perhaps by using quotation marks in your notes from the moment you copy a line. This prevents the common problem in which a borrowed sentence becomes indistinguishable from a self-written note during later drafting.
Step three: build the argument before polishing. Write the main claim and explain what each paragraph must contribute. If a source paragraph looks impressive but does not answer the task, leave it out. Students often over-copy when they have not yet formed a question that makes source information useful. A clear argument gives the student a reason to select, interpret, and challenge evidence rather than collect attractive sentences.
Step four: distinguish quotation, paraphrase, and general knowledge. Exact distinctive language needs appropriate quotation and attribution. Borrowed ideas require acknowledgement even when rephrased. Established disciplinary facts may not always need citations in a school assignment, depending on the task, but controversial, specialised, or surprising claims generally deserve traceable evidence. When in doubt, follow the teacher’s rubric and learn the relevant subject convention.
Step five: explain every source you used. Can you say why this evidence belongs in the paragraph and what it does not show? If not, return to the original source. A citation to a page you have not read creates the appearance of research without the verification. One reliable source understood carefully often helps more than ten references pasted into a bibliography.
Step six: review the similarity report as feedback when access is provided. Open the matches, classify them, and correct genuine attribution or wording problems. Do not aim for a magical percentage or hide proper citations to push the score down. If you cannot access the report, you can still perform the same provenance review using notes, highlighted quotations, and the assignment requirements.
Step seven: preserve evidence of revision responsibly. Save drafts or outlines when the school permits, and make required assistance declarations. This is not a demand to record every keystroke; it is a practical way to remember your decisions and explain how your writing developed. Respect privacy. Do not upload classmates’ work, unpublished school materials, or personal records to unapproved third-party checkers merely because they advertise a free score.
A parent route: move from fear to evidence
If your child comes home upset about a similarity score, begin with three documents: the actual submission, the matched passages, and the assignment instructions. Ask what the child was asked to produce and what they understand about the highlighted sections. Avoid asking them to guess a motive for every match. A student may be genuinely confused by standard quotations or a shared template; another may be embarrassed to admit that they copied a paragraph because they did not know how to paraphrase it. A calm, specific conversation is more informative than either immediate exoneration or immediate punishment.
Parents can help without writing the essay. Choose one highlighted sentence and ask, “Whose information is this? How do you know it is accurate? What does your own paragraph add?” If the child can articulate the relationship between source, evidence, and claim, they are practising the very comprehension skills an English assignment should develop. If they cannot, your next step is to help them locate and read the source, not to invent synonyms. Encourage them to ask the teacher which corrections are permitted and how any concern will be reviewed.
If a formal allegation arises, keep discussion factual. Request the relevant school procedure, understand the evidence to be considered, and allow the student an opportunity to explain the writing process. Do not circulate screenshots of a private report to classmates or online groups. An integrity discussion deserves the same care with privacy and accuracy that students are being asked to show in their research.
A teacher route: make the assessment more diagnostic
A strong teacher workflow begins before the assignment. Teach examples of acceptable quotation, thoughtful paraphrase, unattributed copying, and common factual statements using a short model passage. Make the permitted level of collaboration and technological assistance explicit. If students are expected to revise drafts, specify whether earlier submissions are likely to appear in similarity comparisons. The more the expectations are visible, the less students must reverse-engineer them from a score afterwards.
When reports arrive, scan by match pattern rather than rank order. Check the source type, density, length, and position of substantial matches. Distinguish an administrative exclusion from a substantive dismissal of evidence. If serious concern remains, follow school procedures and give the learner a fair opportunity to respond. Staff should understand whether their institution’s tool retains papers, where matching data comes from, what privacy agreements apply, and whether software output can be audited. Training matters because a confident but unsupported accusation can damage trust as surely as ignoring genuine plagiarism can damage standards.
Teachers can also design assignments that produce richer evidence of thinking: a brief source-choice justification, an annotated paragraph showing why evidence supports a claim, a comparison of two interpretations, or a short oral explanation of a student’s conclusion. These are teaching strategies, not foolproof surveillance techniques. They make learning visible and can reduce overreliance on a single final-file metric. Use them proportionately and accommodate students who need alternative ways to explain their reasoning.
When the simple advice is not enough
Not every matched passage can be understood immediately. A report may associate a student sentence with a site that itself copied the passage from an older source. A student’s earlier work may contain parts produced collaboratively. A long technical description may have little room for stylistic variation. An assignment might include mandatory boilerplate whose exclusion is controlled centrally. Institutional academic-integrity rules may classify some kinds of reuse as unacceptable even when students understand and cite the source. In these cases, follow the actual source and policy rather than a general internet rule.
The tool’s reliability is not uniform across languages, file formats, and source collections. A document converted from PDF may contain broken words or unusual character sequences. Multilingual students may write in conventional patterns taught by a textbook, and that alone is not evidence of misuse. A teacher should avoid inferring motive from imperfect language or from a student who struggles to describe their drafting process under pressure. Similarity evidence can be relevant, but fairness requires triangulation with the task, the document, its sources, and whatever process evidence is appropriate and lawfully available.
There is also an important counterfactual question: would the student’s work satisfy the task if the score were hidden? Does the essay answer the prompt, develop a defensible claim, use appropriate evidence, and acknowledge intellectual debts? If not, the learning problem may be weak argumentation or poor source selection rather than plagiarism. If yes, a few expected matches may not change that judgement. Screening is most useful when it helps teachers find what to examine, not when it replaces examination altogether.
Frequently asked questions
Is 20% similarity a bad score?
Not by itself. The meaning depends on the assignment, the matched passages, source types, settings and school policy. Twenty percent from quotations and required prompts is different from twenty percent of uncited argument copied continuously from a source. Ask what matched and why.
Can an original essay reach 100% similarity?
A paper resubmitted against a stored earlier version of itself could potentially match extensively, depending on the repository and configuration. Such a result still requires review of what was submitted, whether reuse was permitted, and whether the report is identifying the student’s own earlier work. A percentage is not an automatic disciplinary verdict.
Does 0% prove that nothing was copied?
No. Source coverage is incomplete, wording can change, and the reuse of ideas may not resemble the original wording. A low score reports little detected textual overlap under a particular comparison system; it does not certify independent authorship.
Should I replace every highlighted phrase?
No. Leave necessary quotations, correctly cited material, established terminology and appropriate reference details intact. Revise material that copies distinctive language without justification, misrepresents a source, or fails the assignment’s attribution requirements. Make changes to improve scholarship, not merely to lower the metric.
Can I paste my essay into several free checkers to be safe?
Only if school rules and privacy terms permit that. Uploading a student’s original work can expose it to third-party processing or storage and may create future matching complications. Use the school’s approved process where available, and never upload another person’s unpublished work without permission.
Is a similarity score the same as an AI-writing score?
No. Text-matching systems and AI-writing classifiers make different kinds of inferences. Both require careful interpretation and appropriate procedural safeguards. Teachers should not infer unauthorised AI use merely from a similarity percentage.
What if the school already decided based on a number?
Ask which rule was applied, which passages and sources were reviewed, what opportunity the student has to respond, and what review or appeal process exists. Keep the discussion anchored to evidence and policy rather than alleging bad faith. This article is an educational explanation, not legal or disciplinary advice.
Takeaway: let the report open an inquiry, not end one
An academic integrity system should help students learn to distinguish their thinking from the work they draw on. A similarity report can make portions of a document easier to examine, but it cannot tell the full story of writing, citation, intention, or independent understanding. The most reliable practice is pleasantly ordinary: read the assignment, trace the source, inspect the matched words, evaluate the writer’s contribution, and make a fair decision with the necessary context. That skill matters long after a particular school platform changes.
Sources and further reading
- Turnitin Guides — Understanding the similarity score. Provider explanation of the distinction between comparison reports and plagiarism judgements.
- Turnitin Guides — Turnitin and plagiarism. Considerations instructors should review before interpreting matches.
- Turnitin Guides — How exclusion filters refine the Similarity Report. Why quotations and other legitimate material may affect match reports.
- Education Endowment Foundation — Improving Literacy in Secondary Schools. Why reading and writing in each discipline need subject-specific attention.
- Australian Education Research Organisation — Formative assessment practice guide. For turning evidence of student work into better teaching decisions.
Continue exploring eduKateSG
For the wider relationship between learning, explanation and fair assessment, read How Education Works. For research as a learning activity rather than a hunt for quotations, see How School Works. Students working from uncertain claims can also use How Science Works to see why evidence must support the question actually being asked.