Is Scribbr AI Detector Accurate? It Missed 15 of 16 AI Texts

Is Scribbr AI detector accurate? Only on human writing, in our test. Scribbr's free AI Detector scored all 12 of our human texts at 0% AI, but it also scored 12 of our 16 AI texts at 0%, and only one AI text reached 50%. If you paste Claude writing into Scribbr today, it will most likely come back as human.

I've spent nearly 20 years in AI development and digital marketing, and Scribbr is the seventh AI detector I've run through the same 28 texts, after GPTZero, ZeroGPT, Originality.ai, Solvely, Grammarly and JustDone. While testing it on November 10, 2026, I also found out where Scribbr's detector comes from: it is QuillBot's, and an older version than the one on QuillBot's own site.

My Main Points:

• Scribbr flagged none of our 12 human texts. Academic papers, developer docs, personal blogs and essays by English learners all scored 0% AI.
• Scribbr missed 15 of our 16 Claude texts at a 50% cutoff. Twelve scored 0%, and the highest score was 54% for an Opus 5.5 technical doc.
• All four AI texts rewritten by Grammarly's humanizer scored 0% on Scribbr.
• The detector box on Scribbr's page is QuillBot's, labelled v7.1.0. QuillBot's own site runs Model v8.4.0 and scored two of our four shared texts very differently: 0% instead of 54%, and 100% instead of 44%.
• GPTZero, Originality.ai and Solvely caught all 16 of the same AI texts. Scribbr caught one.

Below you'll find every score, how Scribbr compares with the tools in our AI detector accuracy roundup, and when a Scribbr result is worth trusting.

Scribbr AI Detector Results on 28 Texts

Scribbr's free AI Detector got every human text right and almost every AI text wrong on November 10, 2026. The 12 human texts all scored 0% AI. The 16 AI texts averaged 7.8% AI: the Opus 5.5 texts averaged 12.6% and the Fable 5.1 texts 2.9%.

Only four AI texts got any AI score at all. An Opus 5.5 technical doc about environment variables scored 54%, an Opus 5.5 research introduction 44%, a Fable 5.1 technical doc 23% and an Opus 5.5 argument essay 3%. The other 12 AI texts scored 0% AI and 100% human written.

Text group	Texts	Average AI score	Scored 50% or more
Human texts	12	0%	0
Claude Opus 5.5 texts	8	12.6%	1
Claude Fable 5.1 texts	8	2.9%	0
AI texts rewritten by Grammarly's humanizer	4	0%	0

Scribbr's scores don't change from one check to the next. We ran two texts twice: the Opus 5.5 technical doc got 54% both times, and a Fable 5.1 argument essay got 0% both times. That puts Scribbr well ahead of JustDone, whose free detector gave the same text a new random score on every check from our test location, though a steady 0% on AI writing is still a miss.

How We Tested Scribbr's AI Detector

Scribbr got the same 28 texts as every detector in our series: 10 plain AI texts (5 from Claude Opus 5.5, 5 from Claude Fable 5.1), 6 AI texts disguised with a casual voice and typos, and 12 human texts written before 2022. The human set is four peer reviewed introductions from PubMed Central, three developer doc sections for Python, PostgreSQL and Django, two personal blog posts, and three 2006 essays by English learners. Texts run 280 to 390 words.

We used the Free AI Detector on Scribbr's AI Detector page, signed in to a Scribbr account. For each text we pasted it in, clicked Detect AI and saved a screenshot of the result. That is 28 checks, four extra checks on AI texts rewritten by Grammarly's humanizer, and two repeats: 34 checks on November 10, 2026, with no usage limit along the way. We also ran four of the texts through QuillBot's own free detector the same day.

All 16 of our AI texts come from Anthropic's Claude models. Scribbr's page talks about ChatGPT, Gemini and Copilot, so ChatGPT text may score higher than ours did. We kept the same set so the results line up with the six detectors we tested before. We did not test Scribbr's paid Premium AI Detector, Scribbr did not see these results before publication, and no money changed hands.

Scribbr AI Detector Results by Text Type

Scribbr's AI Detector gave its only AI scores to formal writing. The four AI texts it scored above 0% are two technical docs, a research introduction and an argument essay. Every AI blog, explainer and student essay scored 0%, and so did all six AI texts written to sound casual or to include typos.

Scribbr made no mistakes on human writing. The four academic papers, the Python, PostgreSQL and Django docs, the two blogs and the three essays by English learners all scored 0% AI and 100% human written. So a student writing in a second language is unlikely to get a false flag from Scribbr, at least on the evidence of our three essays.

Scribbr splits each result into three bands: AI generated, human written and AI refined, and human written. Scribbr says the middle band spots writing that AI has polished, but it read 0% on all 34 checks, including the four texts Grammarly's humanizer had rewritten.

Scribbr AI Detector vs the Grammarly Humanizer

Scribbr scored all four Grammarly humanizer rewrites at 0% AI. Their originals had scored 44%, 23%, 3% and 0% on Scribbr, so the humanizer took the 44% Opus 5.5 research introduction down to nothing.

On the same four rewrites, Originality.ai and GPTZero caught all four, while Grammarly's own detector and ZeroGPT caught none, as we reported in our Grammarly accuracy test.

Scribbr vs QuillBot AI Detector: Same Engine, Different Versions

Scribbr's AI Detector is QuillBot's detector. On Scribbr's page the detector box loads from quillbot.scribbr.com, the result panel carries a QuillBot logo with the label v7.1.0, and a line under the box says QuillBot's privacy policy and terms of service apply. Scribbr and QuillBot are sister companies: Learneo's own announcement of its LanguageTool deal lists both among its businesses, next to Course Hero and CliffsNotes.

QuillBot's own detector at quillbot.com showed Model v8.4.0 on the same day, and it disagreed with Scribbr on two of the four texts we ran through both.

Text	Scribbr (QuillBot v7.1.0)	quillbot.com (Model v8.4.0)
Opus 5.5 technical doc	54% AI	0% AI
Opus 5.5 research introduction	44% AI	100% AI
Fable 5.1 research introduction	0% AI	0% AI
Human health paper	0% AI	0% AI

Four texts is a small sample, because QuillBot's free version gave us five AI scans for the day without an account, so treat these as spot checks. They still show that Scribbr and QuillBot can score the same text more than 50 points apart, in either direction.

Scribbr itself sends readers to QuillBot for a check it calls "more advance and reliable", and QuillBot's page says its detector is trained to detect content from GPT-6, Gemini, Claude, Llama and more. In our four spot checks QuillBot caught one of three Claude texts, the research introduction Scribbr had scored 44%.

Scribbr AI Detector vs GPTZero, Originality.ai, Grammarly, ZeroGPT and Solvely on the Same Texts

Scribbr caught 1 of our 16 AI texts at 50% or more, the lowest result in our series. GPTZero, Originality.ai and Solvely caught all 16, ZeroGPT caught 10 and Grammarly 8. On the 12 human texts, Scribbr tied GPTZero for the cleanest result: neither gave any human text an AI score above 0%.

Detector	AI texts at 50% or more (of 16)	Average on AI texts	Highest human score	Humanized rewrites caught (of 4)
GPTZero	16	100%	0%	4
Originality.ai	16	93.4%	5%	4
Solvely	16	94.5%	27.2%	not tested
ZeroGPT	10	65.8%	24.9%	0
Grammarly	8	49.8%	15%	0
Scribbr (free)	1	7.8%	0%	0

A detector that calls almost everything human scores perfectly on human texts, so look at the whole set. Count every score of 50% or more as an AI verdict and Scribbr got 13 of 28 texts right, which is 46%. A tool that answered "human" for every text would get 43% on this set.

Scribbr's Accuracy Claims vs Our Results

Scribbr's own accuracy numbers are higher than ours, and they come from different texts. Scribbr's best AI detectors guide, by Koen Driessen and revised July 21, 2026, gives the free detector 78% accuracy and the Premium detector 84%, with no false positives, on 30 texts that include GPT-3.5 and GPT-4 writing.

The AI Detector page is more modest. It says the free detector handles text from GPT2, GPT3 and GPT3.5 with average accuracy, and that the Premium detector is the one that can detect GPT4. The same page's FAQ says the best free tool in Scribbr's research reached 68%, which does not match the 78% in the guide.

None of Scribbr's claims mentions Claude, the model family behind all 16 of our AI texts. Outside tests vary as well: ZDNet's David Gewirtz gave QuillBot's own detector 100% accuracy in October 2025 on five texts, three from ChatGPT and two of his own. Our test covers a different case: the older QuillBot version inside Scribbr, scoring current Claude writing.

When a Scribbr AI Score Helps and When It Doesn't

A Scribbr score of 0% AI does not prove a person wrote the text: 12 of our 16 AI texts got exactly that. Teachers and editors should not use a 0% from Scribbr to clear a piece of writing.

A high Scribbr score is rarer and means more in our data. No human text scored above 0%, and the three AI texts that scored 23% or more were all technical docs or research writing. Even then, treat a score as a reason to ask the writer about their process, which is also what Scribbr's note under the detector says: AI detection alone should never decide someone's academic standing.

In a July 2024 Reddit post, a student wrote that a professor gave their essay a zero after Scribbr's detector rated it 39% to 100% AI, even though their Google Docs history showed them writing it.

Scribbr's habit of missing AI and calling it human also matches the research on detectors in general. A 2023 study of 14 AI detection tools led by Debora Weber-Wulff concluded that the tools were neither accurate nor reliable and leaned toward labelling text as human written instead of catching AI.

For students checking their own drafts, Scribbr is a low risk choice: Scribbr says it is free and unlimited at 1,200 words per check, and in our test it never flagged human writing. For catching AI, GPTZero and Originality.ai did far better on the same texts.

FAQ

Is Scribbr AI detector accurate?

Only on human writing in our November 2026 test. Scribbr's free AI Detector scored all 12 of our human texts at 0% AI, but it also scored 12 of our 16 Claude texts at 0% and gave only one AI text 50% or more.

Does Scribbr AI detector use Turnitin?

The Scribbr AI Detector we tested runs on QuillBot. The detector box loads from quillbot.scribbr.com, shows a QuillBot logo with version v7.1.0, and says QuillBot's privacy policy and terms apply. Scribbr and QuillBot are sister companies owned by Learneo.

Is Scribbr's AI detector the same as QuillBot's?

Scribbr uses QuillBot's detector, but an older version. On November 10, 2026 Scribbr's box showed QuillBot v7.1.0 while quillbot.com showed Model v8.4.0, and the same Opus 5.5 text scored 54% AI on Scribbr and 0% on QuillBot.

Does Scribbr AI detector save your work?

Scribbr says it does not store or share the text you check, and that your text is used only to generate your AI detection report. Because the detector runs on QuillBot, QuillBot's privacy policy also applies, according to the line under the detector box.

Is Scribbr's AI detector free?

Yes. Scribbr says its free AI Detector allows unlimited checks of up to 1,200 words each, and we ran 34 checks in one session without hitting a limit. Scribbr also sells a Premium AI Detector, which we did not test.

Final Thoughts

Is Scribbr AI detector accurate? On human writing it was perfect in our test, and on Claude writing it missed 15 of 16 texts. It flagged none of our 12 human texts, scored 12 of our 16 AI texts at 0%, gave every Grammarly humanizer rewrite 0%, and runs an older QuillBot model than quillbot.com does. Use it to check your own draft if you like, and use GPTZero or Originality.ai when you need to catch AI.

If you need AI tools tested on your own content before you trust them, our custom AI solutions team can run that kind of test for you.