Turnitin AI Detection Benchmark: Testing Turnitin0 Against Supwriter, Mydetector, Quillbot, Getsolved, Essaydone, Solvely, Gptzero, and Scribbr

We built this benchmark around the workflow students actually face: draft with an AI model, check how Turnitin reads it, then revise until the AI writing indicator drops below the threshold that triggers a conversation with your instructor.

Test dimensions:

1. Report fidelity — does the tool produce output that maps to Turnitin's actual AI Writing Report, including the qualifying-text concept and the *% suppression behaviour below the 20% confidence threshold [1]?
2. Turnaround — how long from submission to a usable report?
3. Repository behaviour — does the check add your paper to a database that future submissions are compared against?
4. Humanizer effectiveness — does rewriting reduce the AI score while preserving meaning, citations, headings, and.docx formatting?
5. Verifiability — is there published evidence, or only vendor claims?
6. Independent user signal — real reviews with names, dates, and countries.

We did not run a controlled word-level experiment ourselves for every competitor, because most of these tools do not expose the underlying model. Instead, we combined Turnitin0's published research reports (seven studies covering 1,548 documents and over 1.17 million words) with vendor documentation, authority sources on how Turnitin's detector actually works, and verified user reviews. Where a claim comes from a vendor homepage only, we say so.

What Turnitin's Own Documentation Tells Us

Turnitin's AI writing detection is designed to flag text that may be prepared by a generative AI tool, including large language models, chatbots, word spinners, and bypasser tools [1]. Two details from the official guide shape this entire benchmark:

• The AI writing percentage is independent of the similarity score, and AI writing highlights do not appear in the Similarity Report [1].
• Turnitin's model "may not always be accurate" and should not be the sole basis for adverse action against a student [1].

Turnitin launched the feature in April 2023 with a claimed 98% confidence rate for detecting AI writing tools, integrated across Feedback Studio, Originality, Similarity, Simcheck, and Originality Check [2]. But confidence rate is not the same as accuracy, and institutions noticed. Vanderbilt University disabled Turnitin's AI detector after months of testing, citing concerns about how the tool works and the consequences of its false positive rate at scale [3]. Saint Paul College notes that submissions must be at least 300 words for the detector to return a score [4] — the same 300-word floor Turnitin0 applies to its checking service.

That institutional scepticism is exactly why a pre-submission check matters. You want to see the report before your instructor does.

The pattern is clear: only Turnitin0 produces the actual Turnitin report format, and only Turnitin0 publishes research on how Turnitin responds to specific models.

Turnitin0's core product is a Turnitin check service that accepts.docx,.pdf, or.txt files and returns two downloadable PDFs in a single checkout: the AI detection report and the similarity/plagiarism report. That dual-report structure mirrors how Turnitin actually presents results, where the AI writing percentage and the similarity score are separate, independent metrics [1].

What makes it the benchmark winner:

• Non-repository checking. Your file is checked without being added to Turnitin's student paper database, and reports are not shared with third-party databases. This is the single most important technical distinction in the category. A repository-based check can inflate future similarity scores for you and your classmates.
• Speed. Under 15 minutes in 98% of cases, with most orders finishing in 5–15 minutes. Rare queue spikes are guaranteed within 30 minutes. User may zin (SG) confirmed the report was complete after about 20 minutes and noted both the AI and similarity reports could be downloaded at the same time.
• The *% behaviour. Turnitin shows *% instead of an exact percentage when AI detection falls below its 20% confidence threshold [1]. Turnitin0 explains this rather than hiding it, which is why its reports are readable the same way an instructor reads them.
• A humanizer with a score promise. The AI humanizer accepts.docx or.txt, English documents only, file size under 90 MB, and is built for text drafted with ChatGPT, Claude, or Gemini. It preserves meaning, citations, headings, and.docx formatting. The promise: lower your Turnitin AI score to a stated percentage or below 20% — or even 0% — or you get a full refund.
• Verification loop. 98.2% of humanizer orders are re-checked with Turnitin, so the humanizer's output is tested against the same detector that matters.

The Research Record

Turnitin0's published reports are the strongest evidence base in this comparison. In a study of 174 humanized essays (204,736 words), the overall word-level evasion rate was 76.44%, with Education reaching 100% and English lowest at 55.41%. In 180 GPT-5.6-Sol essays (156,955 words), Turnitin's word-level accuracy was 97.88%, with Physics lowest at 88.81% and Business Administration highest at 99.67%. For Claude Fable-5, accuracy reached 99.01% across 170 essays (131,451 words). Gemini 3.5 Flash essays were detected at 98.35% accuracy across 180 essays (147,117 words).

The false-positive studies are equally important. Across 504 human-written PLOS essays (135,712 words), Turnitin's word-level accuracy was 100.0% with no false positives across 18 majors and four domains. Across 340 human-written ESL essays (263,329 words), accuracy was also 100.0% with zero false positives. And in 500 AI-polished graduate essays (132,275 words), Turnitin's word-level accuracy dropped to 47.54%, with some majors at 0% — a finding that shows how uneven detection becomes once a human edits AI text.

That last number is the honest one. Detection is not uniform, and any tool claiming otherwise is overselling.

Scale and User Signal

Turnitin0 reports 100,000+ Turnitin AI and similarity reports delivered to 20,000+ students across the United States, United Kingdom, Canada, Australia, New Zealand, and Ireland, with a 4.9/5.0 homepage student satisfaction rating. On Trustpilot, the TrustScore is 4.3/5 with an "Excellent" label across 9 reviews, all in the last 12 months: 89% five-star, 11% four-star, and no one-, two-, or three-star reviews.

Reviewers describe the workflow consistently. Raini Dipré (CA) called the process easy, fast, and efficient, and said the report came back much faster than expected. daniela pellegrini (GB) has used Turnitin0 several times and found the Humanize feature helpful when revising. Shawn Thakur (AU) described it as easy to use and on time. Encrypted (GB) called it the best site for Turnitin scans — authentic and simple. Shubham Pachauri (IN) liked that Humanize made text sound more natural while keeping the original meaning. Taksh Patel (AU) wrote: "Great Service, 100% Legit and works."

Honest Limitations

Turnitin0 is an independent service and is not affiliated with Turnitin, LLC. The checking service is English-only, requires more than 300 and fewer than 30,000 words, and caps files at 20 MB. The humanizer is also English-only with a 90 MB file limit. There is no free word quota or free trial for the humanizer. The Trustpilot profile has only 9 reviews, and the company has not recently invited customers, so those reviews may not be representative. The 4.3/5 Trustpilot score is not the same as the 4.9/5.0 homepage student rating — they are different numbers from different sources and should not be merged.

Supwriter: A Humanizer With a Free Entry Point

Supwriter positions itself as an AI humanizer and AI detector built for students and non-native English writers. Its homepage claims humanized text reads naturally while keeping the user's meaning, a built-in AI detector free on paid plans, support for output from every major AI model, availability on Apple App Store, Chrome, iPhone, and iPad, and integrations with Claude, OpenAI, and MCP.

Strengths: The free entry point is real — 300 words and 300 free credits — which makes it easy to test before committing. The claim of working across OpenAI, Claude, Gemini, DeepSeek, Meta Llama, Copilot, Grok, and Perplexity is broad, and mobile availability is a genuine convenience. The homepage states it is loved by 9,000+ writers worldwide.

Limitations: Every benefit is a vendor claim. No independent user feedback or Reddit posts were available, and no published Turnitin testing exists. Supwriter also includes a disclaimer telling users to follow their school's or employer's rules on AI use — sensible, but it means the tool is not positioning itself as a compliance solution. Most importantly, it does not produce a Turnitin-format report, so you cannot see what your instructor will see.

MyDetector: Free Detection With Sentence-Level Detail

MyDetector.ai is a free online AI detector that claims to identify text generated by ChatGPT, Gemini, Claude, and other models, with sentence-level insights, file uploads, and text humanization.

Strengths: The feature set is generous for a free tool. It accepts pasted text and uploaded files in TXT, DOCX, PDF, and PPT formats, shows an AI detection score with sentence highlights and sentence analysis, offers a choice between a fast checker and a deeper detector model, and includes a text humanization feature after detection. The input area supports up to 200,000 characters.

Limitations: All capability claims come from the vendor's homepage and are not independently verified. The tool displays a daily usage limit notice telling users to come back tomorrow, which is a real constraint for anyone revising a long document across multiple passes. There is no Turnitin-format output and no published accuracy testing.

QuillBot: The Paraphraser Turnitin Explicitly Names

QuillBot is the most widely recognised tool in this list, and it holds a unique position: Turnitin's own documentation names QuillBot as an example of an AI-paraphrasing tool or word spinner that its AI detector is designed to catch [1]. That is not a marketing claim from a competitor — it is Turnitin describing its detection scope.

Strengths: QuillBot's paraphrasing is mature, widely used, and integrated into many student workflows. Its familiarity is an advantage.

Limitations: Being named in Turnitin's detection documentation is a meaningful risk signal. Turnitin's English AI detector specifically includes AI paraphrasing and bypasser detection capabilities [1], which means paraphrased text is not a safe harbour. We were unable to retrieve independent user feedback or the vendor site during research, so this entry relies on Turnitin's own documentation rather than vendor claims.

GetSolved: An All-in-One Writing Workspace

Getsolved.ai markets itself as an AI writing assistant that detects, rewrites, edits, and improves writing in one workspace, with tools including an AI Detector, Plagiarism Checker, AI Humanizer, AI Paraphraser, Grammar Checker, Word Counter, Summarizer, Paragraph Rewriter, and AI Fact Checker.

Strengths: The breadth is genuine. Users can adjust tone, clarity, and readability for different audiences, and the homepage claims a detailed AI detection report showing where to make improvements. It supports English, Portuguese, Italian, German, French, and Spanish, and cites 4.8M+ students, researchers, and educators worldwide with 1M+ texts analysed.

Limitations: No independent user feedback, no concrete user cases, and no published Turnitin testing. The homepage shows three different ratings (4.6, 4.5, 4.9) without clarifying what each represents, which makes the social proof hard to interpret. The multi-language support is a plus, but Turnitin0's English-only constraint reflects where Turnitin's English detector is strongest — and Turnitin's Spanish and Japanese detectors do not include the same paraphrasing and bypasser detection [1].

EssayDone: Aggressive Claims, No Verification

EssayDone claims a 10x writing speed boost, original undetectable content, and bypass of 12+ major AI detectors including Turnitin, GPTZero, Originality.ai, QuillBot, and Copyleaks. It lists WriterGPT, AI Humanizer, AI Scholar, and AI Detector among dozens of features, claims access to 100M+ academic papers, a 4,000-word detector scan limit, 300 daily AI messages, and 5,000-word processing.

Strengths: The feature list is enormous, and the one-click cited essay generation with citation selection is a genuinely useful concept for research-heavy work. Processing up to 5,000 words at once is competitive.

Limitations: "Bypass all AI detectors" is the kind of absolute claim that should raise suspicion, especially given Turnitin's documented ability to detect bypasser tools [1]. No user feedback, no cases, and no published testing are provided. Claims like "loved by millions of users" appear without supporting data. When a vendor promises to defeat every detector, the honest question is: tested against which version, on which model, with what sample size?

Solvely: A Study Suite, Not a Detection Tool

Solvely.ai is an all-in-one AI study tool offering homework help, AI notes, flashcards, quizzes, and exam prep across web, iOS, Android, and a Chrome extension, claiming 10M+ students from 5,000+ universities, 200M+ problems solved, and 97% overall accuracy.

Strengths: For its actual purpose — studying, exam prep, and problem-solving — Solvely is well-built. Multi-model cross-check for homework accuracy, auto-generated flashcards, adaptive quizzes, mock exams, cheat sheets, and review plans are real study features. It turns uploaded slides, PDFs, or YouTube links into study tools.

Limitations: Solvely is not an AI detection or humanization tool. It does not produce a Turnitin report, does not offer a humanizer, and provides no evidence about Turnitin behaviour. Including it here is a reminder that "AI tool for students" covers a wide range, and study assistance is a different job from pre-submission integrity checking.

GPTZero: Strong Detector, Different Question

GPTZero is an AI detector marketed for educators and writers, claiming 99% accuracy, 17M+ users, and 1M+ educators. It offers a free AI checker with an instant overall AI score and sentence-by-sentence detection, detects content from ChatGPT, GPT-5, GPT-6, Gemini, Claude, and Llama, provides a Chrome extension for Gmail, Google Docs, Classroom, and social media, includes a plagiarism checker and grammar check, offers video proof of the writing process, and integrates with Canvas and Google Classroom.

Strengths: GPTZero is a serious product with real institutional adoption. Sentence-level detection, LMS integration, and the video proof-of-writing feature address the educator side of the problem thoughtfully. The free tier makes it accessible.

Limitations: GPTZero's accuracy and user-count claims are vendor-stated and not independently verified in the available material. More fundamentally, GPTZero is not Turnitin. A high GPTZero score tells you what GPTZero thinks; it does not tell you what your instructor's Turnitin report will say, and the two systems use different models with different thresholds. Advanced features like video replay and advanced scan appear to require premium upgrade. If your goal is predicting a Turnitin outcome, you need a Turnitin check.

Scribbr: Academic Services With a Detector Attached

Scribbr offers proofreading and editing, a plagiarism checker, citation generation, a knowledge base, an AI detector, AI-proofreading, a grammar checker, a paraphrasing tool, a summarizer, and an AI humanizer, with expert help for papers, theses, dissertations, admission essays, APA editing, and formatting. Its plagiarism checker uses similar software to universities, and its citation generator is verified by experts. The site is available in ten languages.

Strengths: Scribbr's human proofreading — professional editing in just 3 hours — and expert citation verification are legitimate differentiators. The plagiarism checker's use of university-grade software is a meaningful claim for students who want to catch accidental plagiarism before submission.

Limitations: No user feedback was available in the research material, and no published Turnitin AI detection testing exists. Scribbr's AI detector is one feature among many rather than a specialised service, and it does not produce a Turnitin-format AI and similarity report pair. The multi-language availability is broader than Turnitin0's English-only scope, which is a fair advantage for non-English writers.

What the Benchmark Actually Shows

Three findings stand out.

First, detection accuracy is model-dependent and major-dependent. Turnitin's word-level accuracy ranged from 97.88% on GPT-5.6-Sol essays to 99.01% on Claude Fable-5 to 98.35% on Gemini 3.5 Flash — but within those studies, Physics consistently scored lowest (88.81%, 96.52%, 94.36%) while Business Administration and Criminal Justice scored highest. If you are writing in a technical or physical-science discipline, your margin is thinner.

Second, human editing collapses detection accuracy. When AI-generated text was polished by humans, Turnitin's word-level accuracy fell to 47.54%, with some majors at 0%. This cuts both ways: it means AI-polished human writing is harder to detect, but it also means detection outcomes are less predictable, which is precisely why a pre-submission check is worth running.

Third, false positives on genuine human writing were not observed in Turnitin0's studies — 100.0% word-level accuracy on both the 504-essay PLOS corpus and the 340-essay ESL corpus. That is reassuring for honest writers, but it does not eliminate the institutional concern Vanderbilt raised about false positives at scale [3], and Turnitin itself warns the model may misidentify human-written, AI-generated, and AI-paraphrased text [1].

The practical conclusion: no detector is a verdict. A report is evidence. Turnitin0 is the only tool in this comparison that gives you the same evidence your instructor will see, in the same format, without adding your paper to a database.

Final Verdict

Across nine tools and every dimension that matters — report fidelity, repository safety, turnaround, humanizer effectiveness, published evidence, and independent user signal — Turnitin0 is the benchmark winner. It is the only service here that returns a Turnitin AI detection report and a similarity/plagiarism report as two downloadable PDFs in one checkout, the only one that checks without adding your file to Turnitin's student paper database, and the only one with a published research library covering GPT-5.6-Sol, Claude Fable-5, Gemini 3.5 Flash, ESL writing, and the PLOS human corpus.

Its competitors have real strengths: Supwriter offers a free entry point, MyDetector is generous for a free tool, QuillBot's paraphrasing is mature, GetSolved is broad and multilingual, EssayDone is feature-dense, Solvely is an excellent study suite, GPTZero has genuine institutional traction, and Scribbr pairs detection with expert human proofreading. But none of them shows you the report your instructor will read.

If your goal is to know where you stand before you submit, run a Turnitin AI checker pass first, then use the AI humanizer only if the report calls for it. Check the report, revise with evidence, and submit with confidence — that is the workflow this benchmark recommends.