The short answer
Partly. The QuillBot AI detector is free, scores text from 0% to 100% with sentence-level highlights, and QuillBot cites a 99% detection rate on the RAID benchmark. Small independent tests are weaker: Pangram's caught 4 of 9 AI samples. Plagino has not benchmarked QuillBot, so treat its score as a signal.
How accurate is the QuillBot AI detector?
The QuillBot AI detector is very accurate on a public research benchmark and noticeably less accurate in small hands-on tests of newer, edited or short text. Which number applies to you depends on what you paste into it. This section separates QuillBot's own claim from what outside testers found.
Plagino has not benchmarked QuillBot. Our March 2026 benchmark covered seven detectors and QuillBot was not one of them. Every QuillBot figure below is QuillBot's own claim, a public leaderboard result, or a third-party test, and each is labelled.
What QuillBot claims about its accuracy
QuillBot's AI detector page promises "a 99% detection rate, according to independent evaluations from RAID". RAID is a public benchmark for machine-generated text detectors, presented at ACL 2024 by researchers including the University of Pennsylvania team behind it (Dugan et al., 2024). It holds over 6 million generations from 11 models across 8 domains, with 11 adversarial attacks.
We checked the RAID leaderboard on 18 October 2026. QuillBot's entry caught 99.7% of AI text when the threshold was set to allow 5% false positives, and 99.2% at a 1% false positive rate, with no attack applied. Under RAID's paraphrase attack, the 1% figure fell to 94.1%. These are strong results, but RAID's generators include GPT-2, GPT-3, GPT-4 and ChatGPT rather than 2026 models, and its domains include recipes, poetry and Reddit posts as well as academic abstracts.
What independent tests found
Smaller tests on newer text give lower numbers. In a January 2026 test by Pangram Labs, QuillBot's free detector passed all 3 human texts but flagged only 4 of 9 AI samples at a 75% threshold. It caught all three Gemini 2.0 samples, one of three ChatGPT-4o samples and none of the three Claude 3.7 Sonnet samples. Pangram sells a competing detector.
ChatAI Guide's May 2026 mini-benchmark ran 30 English passages written with GPT-5.5-era models. At a 70% threshold, QuillBot flagged 13 of 20 AI or AI-assisted passages and wrongly flagged 1 of 10 human passages. All six long raw AI essays were caught; short paragraphs and paraphrased text were missed most often. Scribbr's own 2026 research, dated 21 July 2026, scored QuillBot's free tool at 78% on 30 short texts, with no human text flagged.
Why the QuillBot accuracy figures disagree
The figures measure different things. RAID reports a detection rate at a fixed false positive rate across millions of mostly older generations. The hands-on tests report a pass count at a chosen threshold on 12 to 30 texts. A detector can score 99% on the first and 44% on the second without either test being wrong.
The practical reading is simple. QuillBot is strongest on long, unedited AI prose and weakest on short or reworked text. QuillBot itself says on its page that "scores are signals, not verdicts."
Which QuillBot accuracy figures can you trust?
Only the RAID leaderboard result and the published hands-on tests can be traced to a method. The table below sorts every QuillBot accuracy figure we found by who produced it and how far it holds up. It is the fastest way to see why one review says 99% and another says 44%.
| Figure | Who produced it | What it measures | How far to trust it |
|---|---|---|---|
| "99% detection rate" | QuillBot, citing RAID | Marketing summary of a benchmark result | Traceable to RAID, but the page does not state the false positive rate it applies at |
| 99.7% at 5% FPR; 99.2% at 1% FPR; 94.1% at 1% FPR under paraphrase attack | RAID public leaderboard, read 18 Oct 2026 | AI text caught at a fixed false positive rate, mostly older models | Strongest evidence available; the entry is submitted by the vendor and the models are pre-2025 |
| 78% overall; 0 human texts flagged | Scribbr research, 21 July 2026 | Closeness-scored accuracy on 30 short GPT-3.5 and GPT-4 texts | Traceable, but tiny, older models, and Scribbr is QuillBot's sister company |
| 4 of 9 AI samples caught (44%) | Pangram Labs, Jan 2026 | Free tool vs GPT-4o, Gemini 2.0 and Claude 3.7 at a 75% pass mark | Traceable, but 12 samples and a competitor's test |
| 13 of 20 AI passages caught; 1 of 10 human passages flagged | ChatAI Guide, May 2026 | 30 English passages at a 70% threshold | Traceable and itemised, but one tester and one session |
| "About 80% accuracy (ZDNet)" | Repeated by GPTZero and other reviews | Said to be a journalist's test | We could not open the original ZDNet article, so we have not verified it |
| "About 13% false positives" | One review citing "aggregated review data" | Unclear | Unverified: no named study behind it |
| 97.4% AI caught, 0.8% human flagged | Plagino's own March 2026 benchmark | Seven tools, 500 AI and 500 human documents | Plagino's figures only; QuillBot was not tested |
Is the QuillBot AI detector free?
Yes. QuillBot's help center says the AI detector "is a free tool available to all users". The free tier allows up to 1,200 words per scan and up to 6 scans per day, and every scan needs at least 80 words.
QuillBot Premium removes the word limit and the daily cap, and unlocks unlimited explainer cards. On QuillBot's premium page, checked 18 October 2026, Premium costs $4.58 a month billed annually. Older reviews still quote $8.33 a month, so confirm the price at checkout. Premium also bundles the paraphraser, grammar checker and a plagiarism checker with 25,000 words a month.
What QuillBot Premium adds to AI detection
The detection model is the same; Premium changes the limits and the explanations. Free users get one explainer card per scan, which says why a passage was flagged. Premium users get explainers across the full text. QuillBot's main page says Premium users can download and share the full report, while its help center says any text of 80 words or more can be downloaded.
How does QuillBot's AI content detector work?
QuillBot's AI content detector estimates how likely a text is to be AI-written by scoring patterns such as predictability, variation in sentence structure and repetitiveness. It returns a percentage from 0% to 100% and highlights the sentences that drove the score. It does not verify authorship.
QuillBot says the detector is trained to recognise text from GPT-5, GPT-4, Claude, Gemini, Llama and other large language models, and that it is retrained as models change. It sorts writing into four classes: AI-generated, AI-generated and AI-refined, human-written and AI-refined, and human-written. The "AI-refined" classes are meant for human writing edited with a tool, which QuillBot's FAQ says reduces false positives on that kind of text. QuillBot says the detector supports 20+ languages; Plagino has not tested that claim.
One design choice matters for anyone checking their own work. QuillBot says that "when the result is unclear, our model tends to classify texts as human written". That lowers false accusations but means a low score is weaker evidence than a high one.
Can QuillBot detect paraphrased or partly AI-written text?
Sometimes. QuillBot says its detector examines text line by line and can flag the AI-written parts of a mixed document. It also says heavily paraphrased or lightly AI-edited content "is harder to detect for any AI checker".
The tests agree. In the ChatAI Guide set, QuillBot caught 3 of 5 lightly hand-edited AI drafts and 2 of 5 paraphrased and reorganised AI texts. Short AI paragraphs of 120 to 180 words fared worst, with 2 of 4 caught. Scribbr's research found QuillBot detected 50% of its mixed and paraphrased texts. On RAID's paraphrase attack the detection rate dropped about five points at a 1% false positive rate.
Will Turnitin catch text that went through QuillBot?
Turnitin says it can. Its AI writing report guide names QuillBot as an example of an AI paraphrasing tool and reports such text as AI-generated and then AI-paraphrased, for English submissions. A clean QuillBot score therefore does not predict a clean Turnitin score. Our guide on whether Turnitin can detect ChatGPT covers what independent tests found.
Does QuillBot's AI detector flag human writing?
It can. QuillBot's FAQ says "all AI checkers present the risk of false positives" and that "highly formulaic content, such as legal jargon or academic writing, is more likely to present a false positive". Its help center says the same: "no tool is perfect."
The evidence points to formulaic text as the risk. In the ChatAI Guide test, the one false positive was a 210-word human technical list scored at 78% AI, while six human essays all stayed under the 70% line. That pattern is not unique to QuillBot. A 2023 Stanford-led study found that several detectors misclassified more than half of TOEFL essays by non-native English speakers as AI-written. QuillBot does not publish a false positive rate for academic essays.
Students who wrote their own work and still got flagged should read what to do if you are falsely accused of using AI. Our guide to AI detector false positives compares the published rates.
Is QuillBot's AI detector the same as Scribbr's?
Not confirmed, but the two are closely linked. QuillBot and Scribbr are both businesses of Learneo, Inc., which acquired QuillBot in August 2021 when the company was still called Course Hero. Scribbr's Spanish-language FAQ says its AI detector works with the QuillBot service.
The results line up too. In Scribbr's research, QuillBot's free detector and Scribbr's free detector both scored 78%. In Pangram's test, both caught 4 of 9 AI samples. Keep that in mind when you read Scribbr's ranking of QuillBot as one of the most accurate free tools: Scribbr is reviewing a sister product. Our Plagino vs Scribbr page covers Scribbr's detector in detail.
How do Plagino and QuillBot compare side by side?
QuillBot is a free, paste-in writing suite with an AI detector built in. Plagino is a dedicated checker that puts AI detection and source matching in one report. Here is how they line up as of October 2026.
| Plagino | QuillBot AI Detector | |
|---|---|---|
| AI docs caught | 97.4% in Plagino's March 2026 benchmark (n=500) | Not tested by Plagino. RAID leaderboard: 99.2% at 1% FPR (older models); hands-on tests: 4 of 9 and 13 of 20 |
| Human docs wrongly flagged | 0.8% in the same benchmark (n=500) | Not published by QuillBot. 0 of 5 (Scribbr), 0 of 3 (Pangram), 1 of 10 (ChatAI Guide) |
| Free tier | 3 scans a month, no card | Up to 1,200 words per scan, up to 6 scans a day |
| Paid price | Pro: $19 a month for 20 scans a day | Premium: $4.58 a month billed annually, no detection word limit |
| Input | Upload a .docx file | Paste text or upload a document; 80-word minimum |
| Report detail | Sentence-level AI highlights and matched sources in one PDF | Sentence-level highlights, four AI/human classes, explainer cards |
| Plagiarism check | Included; matches open-access research through CORE | Separate tool; 25,000 words a month on Premium |
| Languages | Benchmark covered English only | 20+ languages claimed |
Should you use QuillBot or Plagino?
Use QuillBot for a quick, free screen of a draft you can paste. Use Plagino when you want a published false positive rate and an AI check plus source matching in one report. Many students will reasonably use both.
When QuillBot is the better choice
QuillBot wins on cost and convenience. The free tier needs no payment, scans up to 1,200 words six times a day, and takes pasted text. If you already pay for QuillBot Premium for the paraphraser or grammar checker, the detector comes with it at no extra cost.
It also covers more languages. QuillBot claims 20+ languages, while Plagino's benchmark tested English only. And its AI-refined classes are useful if you used a grammar tool on your own sentences and want to see how that reads.
When Plagino is the better choice
Plagino fits students and instructors who need to know the error rate before trusting a score. In our benchmark, Plagino flagged 97.4% of 500 AI-written documents (95% interval roughly 95.6% to 98.5%) and wrongly flagged 0.8% of 500 human-written ones (roughly 0.3% to 2.0%), the lowest false positive rate of the seven tools tested. QuillBot publishes no comparable false positive rate for student essays.
Every Plagino scan also returns matched academic sources alongside the AI highlights, in one PDF you can keep. Pro costs $19 a month for 20 scans a day. You can try Plagino's AI detector free on a draft, or see the page for students.
How should you use a QuillBot AI result?
Use it as one input in a short routine, not as an answer. These steps work for a student checking a draft and for an instructor reading a flagged paper.
- 1
Paste the whole text. Scan at least 300 words where you can. QuillBot says longer texts of 300+ words give more reliable scores than short inputs, and short paragraphs were the weakest category in independent testing.
- 2
Read the highlighted sentences, not just the percentage. Check which sentences drove the score and whether they are formulaic passages such as definitions or method steps.
- 3
Note the class. "Human-written and AI-refined" means the tool thinks a person wrote it and a tool edited it, which is different from AI-generated.
- 4
Get a second reading. Run the same draft through a detector that publishes its false positive rate, such as Plagino, and compare which passages both flag.
- 5
Keep your evidence. Save drafts, notes, sources and version history. If you used AI, disclose it under your course's policy rather than revising to change a score, which is also QuillBot's own advice.
Instructors can see how Plagino fits a marking workflow on the page for teachers.
Test conditions and limitations of this comparison
Plagino's figures come from one benchmark run in March 2026: 1,000 documents (500 AI-written by GPT-4o, Claude and Gemini, some lightly human-edited, and 500 human-written), each tool used on default settings and scored on whether a document was flagged at all. The 95% intervals are Wilson intervals on 500 documents per rate. See the full benchmark method.
QuillBot was not in that run. The RAID leaderboard uses different texts, older models and a fixed false positive rate, so its numbers do not compare directly with Plagino's. The hands-on tests used 12 to 30 samples each. Prices and limits were checked on QuillBot's own pages on 18 October 2026 and change often.
Comparing other tools? See Plagino vs GPTZero, Plagino vs ZeroGPT, Plagino vs Grammarly, or the data-led list of the best AI detectors.
Frequently asked questions
Is the QuillBot AI detector accurate?
It is accurate on long, unedited AI text and less reliable on short or paraphrased text. On the RAID leaderboard it caught 99.2% of AI text at a 1% false positive rate, but small 2026 tests caught 4 of 9 and 13 of 20 AI samples. Treat a score as a prompt to reread, not proof.
Should teachers use QuillBot's AI detector?
Teachers can use it as one screening input, not as a deciding one. QuillBot's own FAQ tells educators to use AI checkers "as one input in a wider conversation with students" and never to rely on detection alone for decisions that affect academic standing.
How much does QuillBot Premium cost?
QuillBot's premium page listed Premium at $4.58 a month billed annually on 18 October 2026. Many reviews still quote $8.33 a month, so check the price at checkout.
What is the minimum word count for QuillBot's AI detector?
QuillBot needs at least 80 words to scan. It says texts of 300 words or more give more reliable scores.
Can QuillBot detect ChatGPT and Claude?
QuillBot says it detects text from GPT-5, GPT-4, ChatGPT, Claude, Gemini and Llama. In Pangram's January 2026 test of its free tool, it caught all three Gemini samples, one of three ChatGPT-4o samples and none of three Claude 3.7 Sonnet samples.
Does QuillBot's AI detector check for plagiarism?
No. QuillBot's FAQ says the AI detector does not scan for plagiarism and points to its separate plagiarism checker. Plagino returns AI highlights and matched sources in the same report.
What does "AI-refined" mean in a QuillBot result?
It means the tool thinks a text was edited by an AI tool rather than written from scratch by one. QuillBot uses it to separate human writing polished with grammar or paraphrasing tools from fully AI-generated text.
Can teachers see if a student used QuillBot?
Not through QuillBot. Teachers only see what their own detector reports, and Turnitin's guide says it flags text changed by AI paraphrasing tools such as QuillBot as AI-paraphrased for English submissions.
Does QuillBot's AI detector work in other languages?
QuillBot says it supports 20+ languages, including English, Spanish, French and Portuguese. Independent tests we found used English only, and Plagino has not tested QuillBot in any language.
Can I download a QuillBot AI detection report?
Yes, for texts of 80 words or more, according to QuillBot's help center. Its main detector page describes downloadable, shareable reports as a Premium feature, so the free option may be limited.
Is QuillBot's AI detector better than GPTZero?
Neither has been shown to be better in a shared, independent test of current models. Plagino benchmarked GPTZero at 88.9% of AI documents caught and 5.7% of human documents flagged, but has not benchmarked QuillBot. See Plagino vs GPTZero for the full comparison.
Will a low QuillBot score mean Turnitin won't flag my paper?
No. Every detector uses its own model, and QuillBot says unclear results lean towards "human". Turnitin's score comes from a different model and is normally shown to instructors rather than students, so a QuillBot result cannot predict it.
