Direct answer
A similarity index is the percentage of a document's text that matches other sources in a database, and a free checker can give you that number — but only against the sources it actually holds. Most free tools index the open web and a limited set of documents, so the figure they return is usually lower than the one your institution will produce [1]. The number itself is not a verdict: it counts quotations, references, and common phrasing alongside genuine copying, which is why a high score can coexist with fully original work [1].
What Does a Similarity Index Actually Measure, and What Percentage Is Considered Acceptable?
The similarity index measures overlap, not misconduct. When a document is compared against a repository, the system highlights every passage that matches a stored source and reports the matched proportion as a percentage of the submitted word count [2]. That means the score is a property of the comparison set, not a fixed quality of your writing: the same essay run against a web-only index and against a full institutional repository can produce two very different figures [2]. A score of 18% against a narrow database may be more concerning than 30% against a broad one, because the first number was produced by a weaker check.
There is no universal acceptable percentage. Turnitin's own guidance is explicit that institutions set their own thresholds and interpret reports in disciplinary context, so a figure that triggers a conversation in one department may be unremarkable in another [2]. What matters to a marker is the pattern of matching — a single unquoted paragraph lifted from a journal article reads very differently from scattered 1% hits on module handbooks, standard definitions, and reference lists [1]. Trying to hit a target number by paraphrasing until the percentage drops is the wrong optimisation, and it tends to damage the argument.
Several mechanical factors inflate the index before any judgement is made. Bibliographies, direct quotations that are correctly cited, appendices, and short common phrases all register as matches unless they are excluded from the calculation [1]. Turnitin reports allow these categories to be filtered out, and a report viewed with exclusions applied can sit many points below the headline figure [1]. If you are comparing your score to a classmate's, confirm you are both looking at the same exclusion settings.
The practical reading is therefore three-layered: check the percentage, then check which sources it came from, then check whether each match is attributed. A high index dominated by your reference list is a formatting issue; a low index containing one unattributed 400-word block is a serious integrity issue [2]. Markers are trained to read reports this way, which is why arguing about the headline number rarely helps.
Are Free Similarity Index Checkers Accurate, and What Do They Miss Compared With Turnitin?
Free checkers are usually accurate at what they claim to do, and that claim is narrower than most students assume. A typical free tool compares your text against publicly crawlable web pages and sometimes a small open corpus, so it will reliably catch a paragraph copied from a well-indexed blog or news site [3]. It will frequently miss subscription journal articles, textbook content, and — most importantly — previously submitted student papers held in a closed repository [3]. Those are precisely the sources that generate matches in coursework.
The consequence is systematic under-reporting rather than random error. Because the missing sources are the ones most likely to match academic writing, a free report tends to return a lower percentage than an institutional check on the same file, and the gap widens for essays that lean on paywalled literature [3]. A clean free report is therefore weak evidence of originality; it confirms only that nothing was lifted from the open web. Students who submit on the strength of a 0% free result and then receive a 34% institutional report are not seeing a bug — they are seeing two different databases.
There is also a data-handling cost that free tools rarely advertise. Several free checkers retain uploaded documents, and some add them to their own comparison index, which means your draft may become a future match for another student's submission [3]. Turnitin's institutional workflow avoids this by design in most configurations, but a third-party site's terms of service are the only thing governing what happens to your file. Before uploading anything to a free checker, read whether the document is stored, for how long, and whether it enters a database.
Finally, most free tools report similarity only. They do not produce an AI writing score, even though AI detection is now a separate and often more consequential line on the same institutional report [3]. A student can pass a free similarity check at 4% and still be flagged for 60% AI-generated content, because the two metrics measure entirely different things and are produced by different models.
How Can I Check My Similarity Index and AI Score Before I Submit?
In most institutional setups you cannot — the report is generated after submission, not before. Turnitin's student guidance describes the report as something you view once your assignment has been submitted to the course and your instructor has made the report available, which by definition happens too late to act on [4]. There is no supported route for a student to run a draft through the institutional repository and receive a pre-submission result, and attempting it through a class submission slot can consume an attempt or register the file permanently [4].
What students actually do is check outside the course, using a service that runs the same comparison engines on a draft. This is where the distinction between a free web crawler and a real report matters most: the value of a pre-submission check comes entirely from whether it sees the sources your institution will see [3]. A check that omits the student paper repository and journal subscriptions reproduces the blind spot you were trying to eliminate. The report you want is the one your professor will open — same similarity percentage, same match list, same AI writing indicator.
Practical workflow for a draft you still have time to change:
- Finish the text, including references, so the check reflects the real document rather than a partial one [1].
- Run the check and read the match list source by source, not just the headline percentage [2].
- Separate legitimate matches (citations, quotations, standard terminology) from unattributed overlap, and fix only the latter.
- Check the AI writing score separately, since similarity and AI detection are independent metrics [3].
- Re-check after revisions if the score moved substantially, because rewriting can shift both figures.
Timing matters as much as accuracy. Reports are most useful when there is still room to rewrite the flagged passages and re-verify the result, which is why a check run the night before a deadline has limited value even if it is perfectly accurate [4]. Build the check into your drafting cycle rather than treating it as a final gate.
If the gap between a free web-only figure and the real institutional report is what worries you, turnitin0 closes it by running your draft through the same Turnitin AI and similarity engines your university uses — non-repository, so your file is not added to the student paper database, with both PDFs delivered together in under 15 minutes in 98% of cases. You see the actual report, with the actual score and match list, while you can still rewrite.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
FAQ
Is a 0% similarity index good?
It means nothing matched the database that was searched, which is not the same as being original. A 0% on a free web-only checker often just reflects a narrow index, and a genuine 0% can also indicate a document that quotes nothing at all [3].
Why is my Turnitin score higher than a free checker's?
Because Turnitin compares against a student paper repository and subscription content that free tools cannot access. The extra sources are the ones academic writing most often matches, so the institutional figure is usually the higher of the two [3].
Should I try to get my similarity index down to zero?
No. Correctly cited quotations and reference lists legitimately register as matches, and stripping them damages the work. Read the match list and fix unattributed overlap rather than chasing a number [1].
Does a similarity check also detect AI writing?
Not usually. Similarity and AI detection are separate metrics produced by separate models, so a clean similarity report tells you nothing about your AI score [3].
Can I see my similarity report before submitting an assignment?
Not through your institution — the report is generated after submission and released by your instructor. Pre-submission checks have to run outside the course workflow [4].