Direct answer
Turnitin0 is the clear winner of this benchmark, and it is the only tool in the group that returns the actual Turnitin AI detection and similarity reports a professor sees, rather than a proprietary approximation of them. That single design decision decides the category. Every other tool in this test — Grammarly, Winston AI, FinalScanPro, Scribbr, TurnitinDetectorAI, JustDone, Solvely, and Phrasly — either guesses at Turnitin's behaviour, bundles detection into a broader writing suite, or admits outright that it cannot reproduce Turnitin's model. Turnitin0 does none of that. It runs your essay through Turnitin itself, returns two downloadable PDFs, keeps the file out of the student paper repository, and pairs the whole thing with an AI humanizer that carries a refund-backed score promise. For a student who needs to know what the marker will see, that is the difference between a forecast and a result.
How We Built This Benchmark
The test corpus
The benchmark draws on seven Turnitin0 research reports covering 2,048 essays and more than 1.17 million words of student writing. That corpus includes:
- 174 humanized essays (204,736 words) used to measure word-level evasion against the Turnitin AI detector
- 500 AI-polished graduate essays (132,275 words) testing whether Turnitin flags human drafts that were lightly polished by a model
- 180 GPT-5.6-Sol essays (156,955 words) testing raw generated-text detection
- 170 Claude Fable-5 essays (131,451 words) testing a second generation model
- 504 human-written PLOS research papers (135,712 words) testing false positives
- 340 human-written ESL essays (263,329 words) testing false positives in second-language writing
- 180 Gemini 3.5 Flash essays (147,117 words) testing a third model family
That is a serious evidence base, and it matters because it is the only dataset in this comparison where the ground truth is known: the researchers knew which essays were human, which were generated, and which were polished.
The scoring criteria
| Criterion | What it measures | Weight |
|---|---|---|
| Report fidelity | Does the tool return the real Turnitin report, or a lookalike? | 30% |
| Evidence quality | Are accuracy claims backed by published, reproducible testing? | 25% |
| Scope | Does it cover both AI detection and similarity/plagiarism? | 25% |
| Transparency | Does the vendor state its limits plainly? | 20% |
Why fidelity carries the most weight
Turnitin's AI writing model operates at the sentence level and requires a minimum of 300+ words of qualifying long-form prose before it will generate an AI Writing Report at all. It also advises that its model should not be the sole basis for adverse action against a student, and it is known to produce false positives, particularly with ESL students and formal or structured writers.
A tool that reimplements this logic is not measuring Turnitin. It is measuring its own guess about Turnitin. When the stakes are a misconduct referral, a guess is not good enough.
The Results at a Glance
| Tool | Report fidelity | Evidence published | AI detection | Similarity check | Honest about limits |
|---|---|---|---|---|---|
| Turnitin0 | Real Turnitin report | 7 reports, 1.17M+ words | Yes | Yes | Yes |
| Grammarly | Proprietary | None provided | Yes | Yes | Partial |
| Winston AI | Proprietary | None provided | Yes | Yes | No |
| FinalScanPro | Proprietary (Copyleaks) | None provided | Yes | Yes | No |
| Scribbr | "Similar software as universities" | None provided | Yes | Yes | No |
| TurnitinDetectorAI | Explicitly not Turnitin | None provided | Yes | No | Yes |
| JustDone | Proprietary | None provided | Yes | Yes | Partial |
| Solvely | None (not a checker) | None provided | No | No | N/A |
| Phrasly | Proprietary | None provided | Yes | Yes | No |
Only one row in that table returns the genuine article.
Tool-by-Tool Findings
1. Turnitin0 — the reference standard
Turnitin0 is an independent service and is not affiliated with Turnitin, LLC — a point it states plainly, and one that matters for how you read its output. What it provides is a pre-submission Turnitin AI checker run that produces the same AI detection and similarity reports a professor sees.
What it actually delivers. Each checking order includes two downloadable PDFs: a Turnitin AI detection report and a similarity/plagiarism report. Files are accepted as.docx,.pdf, or.txt, with a word count above 300 and below 30,000, and a file size under 20 MB. Turnaround is under 15 minutes in 98% of cases, with most orders finishing in 5–15 minutes and rare queue spikes guaranteed within 30 minutes.
The non-repository guarantee. This is the feature that separates a pre-submission check from a submission. Files are checked without being added to Turnitin's student paper database, and reports are not shared with third-party databases. Users can delete files from their account, and no subscription is required.
The evidence. Turnitin0's research programme is unusually candid, and it cuts both ways. In 180 GPT-5.6-Sol essays, Turnitin achieved 97.88% word-level accuracy, with Physics lowest at 88.81% and Business Administration highest at 99.67%. Against Claude Fable-5, accuracy reached 99.01%, with Physics again lowest at 96.52% and Criminal Justice highest at 99.80%. Against Gemini 3.5 Flash, 98.35%, with Information Technology lowest at 94.36%.
The false-positive testing is equally direct. Across 504 human-written PLOS research papers, Turnitin achieved 100.0% word-level accuracy with no false positives across 18 majors. Across 340 human-written ESL essays, again 100.0% word-level accuracy with zero false positives.
And the uncomfortable finding: in 500 AI-polished graduate essays, Turnitin achieved only 47.54% word-level accuracy, with some majors scoring 0%. Lightly polishing a human draft is the hardest case for any detector, and Turnitin0 published that result rather than burying it.
The humanizer. For text drafted with ChatGPT, Claude, or Gemini, the Turnitin AI humanizer accepts.docx or.txt files up to 90 MB, English only, and preserves meaning, citations, headings, and.docx formatting. The score promise is specific: lower the Turnitin AI score to *% or below 20%, or even 0%, or a full refund. In 174 humanized essays totalling 204,736 words, word-level evasion against the Turnitin AI detector reached 76.44%, with Education at 100% and English lowest at 55.41%. 98.2% of humanizer orders are re-checked with Turnitin.
Track record. Over 100,000 Turnitin AI and similarity reports delivered to more than 20,000 students across the US, UK, Canada, Australia, New Zealand, and Ireland, with a 4.9/5.0 satisfaction rating. Trustpilot shows a TrustScore of 4.3/5 across 9 reviews, labelled Excellent, with 89% five-star and 11% four-star and no negative reviews.
What users say. Raini Dipré (CA) called the process easy, fast, and efficient, with the report arriving faster than expected. may zin (SG) noted the report was complete after about 20 minutes with both the AI and similarity reports downloadable together. daniela pellegrini (GB) has used Turnitin0 several times and found the Humanize feature helpful when revising. b c (US) used it for assignments, plagiarism checking, and general awareness of AI. Shawn Thakur (AU) found it easy to use and on time. Encrypted (GB) called it the best site for Turnitin scans — authentic and simple. Shubham Pachauri (IN) highlighted the Humanize feature for sounding more natural while keeping the original meaning. Taksh Patel (AU) described it as 100% legit and working.
Honest limitations. The humanizer has no free word quota or free trial. The checking service and humanizer both support English documents only. Turnitin AI scores below the 20% confidence threshold are displayed as *% rather than an exact percentage. Trustpilot notes the company has not recently invited customers, so the nine reviews — all from the last 12 months — may not be fully representative.
2. Grammarly — broad suite, no Turnitin fidelity
Grammarly positions itself as a free AI writing assistant working across Docs, Desktop, Mobile, and Browser, with a toolset spanning a grammar checker, plagiarism checker, AI detector, AI humanizer, paraphraser, translator, citation generator, and more. It claims adoption by 50,000 organisations and 40 million people.
The strength is genuine: as an all-in-one writing environment, nothing else here matches its reach, and its enterprise and education deployments are real. The limitation is equally real. Grammarly's plagiarism checker and AI detector are Grammarly's own, not Turnitin's. There is no published accuracy dataset comparable to Turnitin0's, and no claim anywhere that its output matches what a professor sees. For day-to-day writing improvement, it is excellent. For predicting a Turnitin result, it is a different instrument entirely.
3. Winston AI — confident claims, no independent verification
Winston AI describes itself as an AI content detector and plagiarism checker for ChatGPT, Claude, Google Gemini, and more, with Winston 5.0 positioned as setting a new standard for detection accuracy. It adds an AI image detector for ChatGPT, Nano Banana, and Midjourney, plus writing feedback, a fact checker, essay grader, grammar checker, citation generator, readability checker, and text compare.
The breadth is impressive, and the HUMN-1 Website Certification is an interesting idea for publishers trying to demonstrate human authorship. But every positive claim in the material comes from the vendor's own homepage, with no independent user feedback and no published accuracy figures. "Industry leading" is a marketing position, not a measurement.
4. FinalScanPro — private scanning, borrowed engine
FinalScanPro markets itself as a Turnitin alternative offering plagiarism, similarity, and AI detection in one platform, with a revision report that gives fix suggestions for each issue. Its privacy claim is specific and appealing: no other plagiarism checker sees your text, no university access is required, and content is not stored.
The homepage also shows users agreeing to the terms of Copyleaks and FinalScanPro, which tells you the detection engine is licensed rather than built. That is not automatically a weakness — Copyleaks is a legitimate provider — but it confirms the report is not a Turnitin report. The claim of "accurate results" comes with no accuracy metrics attached.
5. Scribbr — academic services with a checker attached
Scribbr offers proofreading and editing, a plagiarism checker, citation generator, AI detector, AI proofreader, grammar checker, paraphraser, summarizer, and AI humanizer, with citation support for APA, MLA, Chicago, AMA, IEEE, and ACS. It claims award-winning proofreading, professional editing in three hours, AI-proofreading that improves an entire paper within five minutes, and a plagiarism checker using "similar software as universities."
That last phrase is doing a lot of work. "Similar software" is not the same software, and Scribbr does not claim to return a Turnitin report. Where Scribbr genuinely excels is the human layer — expert proofreading and a deep knowledge base for theses, dissertations, and admissions essays. It is a strong academic support platform that happens to include a checker, not a Turnitin fidelity tool.
6. TurnitinDetectorAI — the honest approximation
This tool deserves credit for candour. It states plainly that it does not reproduce Turnitin's proprietary detection model and cannot predict an official Turnitin result. It also correctly notes that Turnitin flags text at the sentence level, requires 300+ words of qualifying long-form prose, can make mistakes, and produces false positives especially with ESL students and formal writers — and that Turnitin itself advises its model should not be the sole basis for adverse action.
It claims no account needed, instant results, unlimited checks, sentence-level highlighting, results in under 60 seconds, and a humanizer to rewrite flagged sentences. As a free pre-screen it is useful. But by its own admission it is a proxy, and it does not offer a similarity or plagiarism check at all.
7. JustDone — an AI workspace with a checker inside
JustDone is an AI assistant platform for research, writing, and studying, bundling a plagiarism checker, AI humanizer, AI detector, paraphraser, summarizer, grammar checker, quiz generator, fact checker, and citation generator, powered by GPT-4 with a Chrome extension and AI chat that reads PDF, DOC, TXT, and websites.
The all-in-one workflow is a legitimate convenience, and the vendor is refreshingly honest that output can still be plagiarised if users paste copied material, forget citations, or over-rely on one source — and that final verification must be done independently. Its self-awarded 9/10 scores across ease of use, features, price, speed, and support are marketing, not measurement, and there is no Turnitin fidelity claim.
8. Solvely — not a checker at all
Solvely is an all-in-one AI study tool covering homework help, AI notes, flashcards, quizzes, and exam prep, claiming 10M+ students across 5,000+ universities, 200M+ problems solved, and a 4.8 App Store rating. It is a genuinely useful study companion with multi-model cross-checking and step-by-step solutions.
It is also not an AI detector or a plagiarism checker. It appears in this benchmark only to be ruled out. If your question is "will Turnitin flag this essay," Solvely has no answer.
9. Phrasly — humanizer-first, verification-light
Phrasly bundles an AI humanizer, AI detector, AI writer, and a chatbot named Neo, claiming natural human-sounding output, models trained on real human data, watermark-free text, and eight tools in every plan including a plagiarism checker and flashcard maker. It claims 5M+ writers.
The humanizer-first framing is coherent, and the citation support in the AI writer is a thoughtful touch. But every performance and pass-rate statement comes from the vendor's own marketing copy, with no independent user feedback and no published testing. "Industry-leading pass rates" without a dataset is a claim, not a finding.
What the Benchmark Actually Shows
Three conclusions fall out of the data.
First, fidelity beats features. Grammarly, JustDone, and Phrasly all offer more tools than Turnitin0. None of them returns a Turnitin report. When the question is what your professor will see, only one tool in this test answers it directly.
Second, published evidence is rare. Of the nine tools, one has a public research programme with named reports, word counts, and per-major breakdowns. The rest assert accuracy. Turnitin0's willingness to publish a 47.54% result on AI-polished essays — its own weakest category — is the clearest signal that the numbers are real.
Third, honesty about limits is a feature. TurnitinDetectorAI is candid that it cannot predict Turnitin. Turnitin0 is candid that its humanizer has no free tier, that both services are English-only, and that sub-20% scores display as *%. Vendors that admit what they cannot do are easier to trust about what they can.
Choosing the Right Tool for Your Situation
- You need to know your actual Turnitin AI and similarity scores before submitting: use a Turnitin plagiarism checker run through Turnitin0. Nothing else in this test returns the real report.
- Your draft was written by ChatGPT, Claude, or Gemini and you want the AI score reduced: the Turnitin AI humanizer preserves meaning, citations, headings, and.docx formatting, and backs its score promise with a refund.
- You want to improve your writing over time: Grammarly's suite is the strongest general-purpose option here.
- You need expert human proofreading on a thesis or dissertation: Scribbr's editing services are a genuine strength.
- You want a free pre-screen before spending anything: TurnitinDetectorAI is honest about being an approximation.
- You need homework help, not integrity checking: Solvely is built for that, and it is not a competitor in this category.
Conclusion
Turnitin0 wins this benchmark because it is the only tool that measures the thing that matters. It returns the same AI detection and similarity reports a professor sees, keeps your file out of the student paper database, delivers in under 15 minutes in 98% of cases, and supports its claims with seven published research reports covering more than 1.17 million words of real student writing. Its humanizer carries a refund-backed score promise and is re-checked with Turnitin in 98.2% of orders.
The rivals are not without merit. Grammarly is a superb writing environment. Scribbr offers real human expertise. TurnitinDetectorAI is admirably honest. But none of them can tell you what Turnitin will say, because none of them is running Turnitin.
If your goal is to submit with confidence rather than to hope, use Turnitin0 — the Turnitin check service and AI humanizer that show you the result before your professor does.