Guide

Copyleaks AI Detector: How It Works, How Accurate It Is, and What It Misses

A plain-English guide to the Copyleaks AI Detector: what its published accuracy figures measure, what independent tests found, how the sensitivity levels change a score, and what AI Logic and AI Source Match do.

September 28, 2026 · Updated November 5, 2026 · The AI Rankings

Quick answer: The Copyleaks AI Detector is a commercial AI-text detector sold to schools, universities and businesses, with a free web version (Copyleaks). Copyleaks advertises over 99% accuracy and publishes a dated testing methodology behind it. The methodology documents model V11, tested on 10 September 2026, and reports a 0.30% false-positive rate at the default Balanced sensitivity (Copyleaks testing methodology). In a peer-reviewed study published on 29 June 2026, Copyleaks cleared every human-written paper with no false positives, and gave papers written entirely by AI an average AI score of 14% (Van Vlasselaer, Van Droogenbroeck and Spruyt, International Journal for Educational Integrity). That study tested an earlier Copyleaks model on long academic papers written with GPT-4o Deep Research. Copyleaks’ own guidance is that detection works best “when paired with human judgment”.

This guide covers the Copyleaks AI Detector specifically. For the underlying detection mechanism across all tools, see how AI detectors work. For ranked alternatives, see best AI detectors. For the institutional detector most students are assessed by, see Turnitin AI detection. Copyleaks also sells plagiarism detection, a separate product covered in best plagiarism checkers.

What the Copyleaks AI detector is

Copyleaks began as a plagiarism-detection company and added AI-text detection later. Four facts about the product shape everything below.

It is an institutional product with a free web version. Copyleaks sells to universities, schools, publishers and enterprises, and integrates with learning management systems. Anyone can also paste text into the web detector and get a percentage back (Copyleaks).

It reports a score plus an explanation. Copyleaks calls this layer AI Logic, and it has two parts: AI Phrases, which highlights the sentences whose patterns drove the score, and AI Source Match, which flags where text matches content already published on AI-generated websites (Copyleaks).

Detection sensitivity is a setting. Copyleaks offers three detection levels: level 1 catches text copied straight from a model with no edits, level 2 catches text with minor changes such as tense adjustments or added words, and level 3 catches heavily modified AI text (St Petersburg College faculty documentation). The same document says that at that college the Canvas default is level 2, and that an instructor can only change the setting before students submit.

It is one product in a suite. Copyleaks also sells AI image detection, a plagiarism checker and Codeleaks for AI-generated and plagiarised source code. The accuracy figures on this page concern the text detector only.

How Copyleaks’ detector works

It classifies per section and aggregates. The API returns “per-section classification (human or AI), an overall human-versus-AI summary, and AI Logic explanations” (Copyleaks API documentation). The web product shows this as a document-level percentage with AI-attributed phrases highlighted.

AI Phrases identifies sentences whose linguistic patterns are statistically characteristic of model output, the same broad kind of signal other classifier-based detectors use. The mechanism is set out in how AI detectors work.

AI Source Match checks flagged passages against content already published on AI-generated websites. When it finds a match, it returns a named external source that can be checked. When it finds none, that says nothing either way about the text.

Copyleaks says it detects modified text. It says the detector identifies “mixed text where human-created text has been amended with AI-generated text” and “paraphrased content, including text that has been processed through spinners or modified with deliberate typos” (Copyleaks AI Detector FAQs), and the API documentation adds detection of character swaps and hidden characters.

Version history is public. Copyleaks shipped AI Text Detection model V11.0 on 22 September 2026, saying the release “improves detection accuracy and extends coverage to recently released large language models”, and API responses report the model version as modelVersion: v11.0 (Copyleaks release notes). On 20 October 2026 it released dedicated detection models for German, Spanish, French, Italian and Portuguese.

What Copyleaks will and will not scan

Short text is outside the tested range. Copyleaks will return a percentage from 255 characters, roughly 40 words. Its published test used texts above 350 characters, and the 2026 peer-reviewed study used full academic papers, so a score on a single paragraph has not been measured by either.

Explanations cover fewer languages than detection. Copyleaks detects AI text in 30 languages and offers AI Logic phrase-level explanations in six. Every independent study cited on this page tested English.

How to read a Copyleaks AI score

A percentage is not a proportion of your document. It is an aggregate of section-level classifications.

The same text can score differently at different sensitivity levels. Because the level is an institutional or instructor setting, the same essay can produce different results at two institutions. If a score is being quoted to you, it is reasonable to ask which level it was produced at.

The AI score and the similarity score are separate numbers. Copyleaks states that the similarity score shows matching against online sources, while the AI percentage estimates generative-AI involvement (Copyleaks AI Detector FAQs). A 0% similarity score and an 80% AI score are not contradictory.

The AI Phrases show what drove the score. The highlighted sentences are the part of a result you can examine directly.

Which AI models Copyleaks says it can detect

Copyleaks names ChatGPT, Gemini, Claude, DeepSeek and Llama on its detector page, and its V11 QA evaluation covered text from OpenAI, Claude, Gemini and Grok model families (Copyleaks testing methodology). Copyleaks reports aggregate rates across its whole AI test set rather than a figure for each model.

What Copyleaks says about accuracy

Copyleaks publishes a testing methodology with results from two separate teams (Copyleaks testing methodology). The current page documents model V11, tested on 10 September 2026 and published on 28 September 2026.

Copyleaks’ FAQ describes false positives as “rare (0.2%)” (Copyleaks AI Detector FAQs). Its FAQ also notes that studies show human-written content, “especially from non-native English speakers”, can be mistakenly flagged by AI detectors, and Copyleaks reports 99.84% accuracy on its own non-native English test sets.

These are vendor figures, measured on Copyleaks’ own test sets. Copyleaks publishes the counts and minimum length of those sets, but not the length distribution, the prompts or the share of edited text.

Copyleaks’ list of third-party studies

Copyleaks maintains a page listing 17 items it describes as third-party studies of its detector, dated 8 July 2025 and updated since (Copyleaks). By Copyleaks’ own descriptions, they include:

The named samples range from four documents to a few hundred. The list includes Alshammari and Rao’s March 2025 study, which reported Copyleaks at 100% accuracy on human text and 99.7% on DeepSeek DeepThink output. The 2026 Vrije Universiteit Brussel study below is not on the list.

What independent testing found

The 2026 Vrije Universiteit Brussel study tested GPTZero, Copyleaks, Turnitin and Pangram (VUB). On the 40 human-written papers, Copyleaks returned 0.0% AI on average, with no false positives. On the 40 papers written entirely with GPT-4o Deep Research, its average AI score was 14.0%, underestimating the AI content by a mean of 85.97 percentage points. GPTZero’s result on those papers was almost identical. The authors suggest the Deep Research model may explain why most of the tools did less well on fully AI papers than on hybrid ones. The study tested an earlier Copyleaks model than the current V11. Its authors concluded that detection tools “should not be used as sole evidence in high-stakes decision-making but should be implemented in a broader evaluation strategy.”

The Kar et al. study used a single AI-written article of under 500 words. Copyleaks scored the original and two paraphrased versions at 100% AI, and a QuillBot-paraphrased version at 0% (Kar et al.). It is one sample, so it shows a single result rather than a rate.

One clarification. The widely cited 2023 study by Weber-Wulff and colleagues, which tested 14 detectors, did not test Copyleaks (Weber-Wulff et al.).

False positives and a reported case

On the evidence above, Copyleaks rarely flags human writing: its own V11 test reports 0.30% false positives at the default setting, and the 2026 study found none across 40 human papers. The trade-off for any detector tuned this way is that it can miss AI-written text, which is what the 2026 study measured on fully AI papers.

A reported case. EdSurge reported in April 2024 that Marley Stevens, a student at the University of North Georgia, was accused of cheating; she says she used only Grammarly’s grammar and spell-check features. Turnitin flagged the paper, and her professor said Copyleaks also flagged it. When Stevens later ran the paper through Copyleaks herself, it came back as human-written. Copyleaks did not respond to EdSurge’s request for comment (EdSurge, 4 April 2024). For what grammar tools do to text, see best AI grammar checkers.

Copyleaks in the LMS

It runs inside the major learning management systems. Copyleaks launched AI Logic across Canvas, D2L Brightspace, Moodle, Blackboard, Schoology, Edsby and Sakai on 29 July 2025 (Copyleaks). Where an institution has it enabled, a submission can be scanned on upload and the result appears in the instructor’s view.

Copyleaks’ guidance is that the score informs a conversation. Its FAQ says “data provided by AI detectors should be used to inform the situation and offer the option for a learning opportunity”, that detection works best “when paired with human judgment and used as part of a broader content evaluation process”, and that no tool should be used in isolation for high-stakes decisions.

Copyleaks says it serves millions of users across enterprise and education and has appeared on the Inc. 5000 list.

Plagiarism checking and Codeleaks

The plagiarism checker is a separate product. It matches submitted text against published sources and produces a similarity score, a different number from the AI percentage. For that comparison, see best plagiarism checkers.

Codeleaks is the source-code product. It detects AI-generated and plagiarised code. The one study Copyleaks cites on code detection is a student thesis reporting that it exceeded an 80% accuracy requirement. For reviewing AI-written code on its merits, see best AI code review.

Image detection gained confidence heatmaps on 23 September 2026 (Copyleaks release notes). For image detection generally, see best AI image detectors.

What it costs

Prices are in USD (Copyleaks pricing). Credits are shared across text and image scanning. Annual billing gives the same credits per month at a lower price: about 18% less on Personal and 25% less on Pro.

What to do if Copyleaks flags your work

If you are a student. Copyleaks’ own guidance says the score should inform a conversation rather than settle one. The strongest material you have is process evidence: version history from Google Docs or Word, timestamped drafts, outlines and notes. Ask which sensitivity level the scan ran at, and whether AI Source Match returned a named source or only a classifier score.

If you are a teacher or administrator. Follow Copyleaks’ guidance to pair the result with human judgment. Best AI for teachers covers assessment approaches that do not depend on detection.

If you are checking your own writing. The free scan is enough for a spot check. Read the AI Phrases rather than the percentage.

If you are deciding whether to pay. Test the free tier on documents whose origin you already know, including a long AI-written piece, before buying.

Who runs Copyleaks

Copyleaks, Inc. was co-founded by Alon Yamin, its chief executive, and Yehonatan Bitton, its chief technology officer. It raised a $6 million Series A in April 2022 led by JAL Ventures with participation from Connecticut Innovations, bringing total disclosed funding to $7.8 million including a $1.8 million seed round. At that point the company had around 20 full-time staff, roughly half in Israel (Calcalist, 19 April 2022).

Frequently asked questions

Is Copyleaks accurate?

Copyleaks advertises over 99% accuracy. Its published test of model V11, run on 10 September 2026, reports 0.9970 accuracy on human text and 0.9936 on AI text, with 0.30% false positives at the default setting, measured on its own test set. In a peer-reviewed study published on 29 June 2026, an earlier Copyleaks model produced no false positives on 40 human-written papers and gave 40 fully AI-written papers an average AI score of 14%.

Does Copyleaks detect ChatGPT?

Copyleaks says so: it names ChatGPT, Gemini, Claude, DeepSeek and Llama on its detector page, and its V11 test set included output from OpenAI, Claude, Gemini and Grok model families. A 2024 study in the Indian Journal of Psychological Medicine scored an unedited ChatGPT-written article at 100% AI. Copyleaks does not publish a detection rate for each model.

Is Copyleaks free?

Partly. The web detector has a free version with usage limits. The Personal plan is $16.99 per month billed monthly, or $13.99 per month billed annually; Pro is $99.99 monthly or $74.99 per month billed annually. Education and Enterprise plans are quoted on request.

Why does Copyleaks say my writing is AI when I wrote it myself?

Copyleaks reports a 0.30% false-positive rate at its default setting on its own test set, and the 2026 peer-reviewed study found no false positives across 40 human papers. When a detector does flag human writing, it is responding to how predictable and regular the prose is, not to who wrote it. Copyleaks’ FAQ notes that studies show writing by non-native English speakers can be mistakenly flagged by detectors. The sensitivity level your institution chose can also change the result.

Does Copyleaks detect paraphrased or humanised text?

Copyleaks says it detects paraphrased text, including text run through spinners. In the 2024 Indian Journal of Psychological Medicine study, it scored an AI-written article and two paraphrased versions at 100% AI, and a QuillBot-paraphrased version of the same text at 0%. That was a single sample of under 500 words.

What does the Copyleaks AI percentage mean?

It is an aggregate of section-level classifications, not the proportion of your document that came from a chatbot. It is separate from the Copyleaks similarity score, which measures matching against published sources. It also depends on which of the three sensitivity levels the scan ran at.

What are AI Phrases and AI Source Match?

They are the two parts of Copyleaks’ AI Logic layer. AI Phrases highlights the sentences whose patterns drove the score. AI Source Match checks flagged text against content already published on AI-generated websites and can return a named external source. AI Logic runs across Canvas, D2L Brightspace, Moodle, Blackboard, Schoology, Edsby and Sakai.

How many words does Copyleaks need?

Copyleaks sets a minimum of 255 characters on the web platform and 350 characters in the browser extension, roughly 40 to 55 words. Its published test used texts above 350 characters, so very short passages are outside the tested range.

Do universities use Copyleaks?

Yes. Copyleaks integrates with Canvas, D2L Brightspace, Moodle, Blackboard, Schoology, Edsby and Sakai, and where an institution enables it a submission can be scanned automatically with the result shown to the instructor. Copyleaks’ guidance is that results “should be used to inform the situation and offer the option for a learning opportunity”.

Who owns Copyleaks?

Copyleaks, Inc. was co-founded by Alon Yamin, its chief executive, and Yehonatan Bitton, its chief technology officer. Its disclosed funding is $7.8 million, from a $1.8 million seed round and a $6 million Series A announced in April 2022.

Is Copyleaks the same as its plagiarism checker?

No. The plagiarism checker matches text against published sources and produces a similarity score; the AI detector estimates generative-AI involvement and produces an AI percentage. Copyleaks also sells Codeleaks for source code and an AI image detector.

← All guides