CopyLeaks Review: The AI Content Detector For Educators
Responsible use of technology is more important now than ever before as our reliance on AI continues to grow. Copyleaks is an AI content and plagiarism detection tool that’s meant to aid teachers in ensuring academic integrity. But how effective is it really?
John Angelo Yap
Updated September 21, 2023
Robot Pretending To Be A Student
Reading Time: 8 minutes
Across history, there have been many inventions made to address issues caused by earlier inventions. The invention of cars caused automobile accidents, so seat belts were created. Single-use plastics caused pollution, so biodegradable packaging was invented. And so on and so forth…
Everyone was aware that AI would have consequences — but, even at this relatively early stage, none were prepared at the scale of its implications. Due to its accessibility, bad actors have exploited tools like ChatGPT for their own malicious intent. This is especially prevalent in schools, where students have used it to cheat on their schoolwork.
The invention made to address issues from AI chatbots? AI content detectors. These are specialized software made to analyze texts word-per-word to identify possible instances of AI.
There’s Content At Scale, Originality AI, GPTZero, and the always unreliable but improving TurnItIn. However, there’s one name that stands out due to its supposed accuracy: Copyleaks.
The question is, how accurate is it really? Find out more in this article.
What is Copyleaks?
Copyleaks is an analysis tool for AI and plagiarism detection. It uses advanced artificial intelligence algorithms to scan text, code, and even images and gauge content authenticity and integrity.
Furthermore, it also has a history that’s just as interesting as its feature base. Copyleaks was first introduced to the world in 2015 by Alon Yamin and Yehonatan Bitton — two ex-military men from Israel whose time in the service were mostly spent on text analysis technology. In fact, this was where these two met and decided to build the award-winning platform we’ve come to know as Copyleaks.
Who Should Use Copyleaks?
Copyleaks is the perfect platform for educators to ensure student integrity in the age of artificial intelligence. Not only does it act as a filter for AI content, it also has other useful features for teachers everywhere like plagiarism detection, automatic grader, and text comparison. Copyleaks also streamlines cheating detection through direct integration with popular LMS like Canvas, Blackboard, and Moodle.
Apart from the education industry, Copyleaks is also a great enterprise solution for monitoring content that could be flagged as plagiarized or AI-generated through a feature called “Gen AI Governance,” which will be discussed further in later sections.
Copyleaks Key Features
AI Detection
Copyleaks is most known for its robust detection tool that can classify if a text came from ChatGPT or other popular generative AI software. Once you enter a text, it only takes a few seconds until Copyleaks can tell you whether it came from an AI or not. As a testament to its accuracy, an independent study in July 2023 found Copyleaks to be the most reliable AI text classifier at 99.12%.
The best thing about their AI detection tool is that — unlike the rest of their features — it’s completely free.
Plagiarism Detection
Apart from being an AI detection tool, Copyleaks is most known for their plagiarism detector. All you have to do is upload a piece of text, document, or even code into their dashboard and wait for a moment. Once Copyleaks is done scanning your document, you’ll get access to information like how much of the paper is similar to other content, word similarities, and which pieces were identical or paraphrased.
Generative AI Governance
In Copyleaks, the buck doesn’t stop at AI and plagiarism detection — they’re fully committed to ensuring internet-wide content authenticity. That’s the reason for Gen AI Governance: a suite of tools to help companies manage and govern the use of generative AI in content creation.
This feature is divided into two:
- Auditing: Checks your AI-generated content, including code and images, and ensures it complies with industry standards.
- Monitoring: Real-time processing of content to limit content generated by AI tools.
AI-Powered Grader
Primarily for educators, Copyleaks’ AI-powered grader is designed to provide accurate assessment evaluations by bulk processing. This is a three-step process that starts with an AI analyzing previously graded sheets to understand how they’re marked. After that, the grader emulates how papers were ranked and the process ends with the students’ grades as an output.
That said, skepticism is healthy and — with tall promises — I won’t be surprised if you’re wary of using Copyleaks as a grader. As a proof of precision, Copyleaks teamed up with the Ministry of Education in Israel to test the algorithm. The result was a huge success with only a 1% margin of error against human graders.
Other Notable Features
- API and LMS Integration: Streamlines your AI and plagiarism detection through integration with different applications, websites, and learning management systems.
- Multilingual Processing: Copyleaks can process text from 100+ languages.
- Text Comparison: Compares two pieces of text to eliminate the possibility of copied assessments.
How Accurate is Copyleaks' AI Detector?
Copyleaks proudly states that they have a 99.12% AI detection accuracy. That’s a bold claim. Let’s put that to a test, shall we?
ChatGPT
Copyleaks proudly states that they have a 99.12% AI detection accuracy. That’s a bold claim. Let’s put that to a test, shall we?
ChatGPT Test #1: Essay
TEST SUCCESSFUL! — Likelihood of AI Content: 96.5%
ChatGPT Test #2: Blog Post
TEST SUCCESSFUL! — Likelihood of AI Content: 75.7%
ChatGPT Test #3: Cover Letter
TEST SUCCESSFUL! — Likelihood of AI Content: 83.6%
ChatGPT Test #4: Prose
TEST SUCCESSFUL! — Likelihood of AI Content: 98.2%
Modified ChatGPT Text (AI Bypass Tools)
After detection tools became more popular, people started looking for ways to rewrite their AI-generated content to avoid getting flagged. For this exercise, I’ll be using two of my favorite AI bypass tools: Undetectable AI and HideMyAI.
Here’s how well they evade Copyleaks detection:
Undetectable AI Test #1: Essay
TEST SUCCESSFUL! — Likelihood of AI Content: 86.3%
Undetectable AI Test #2: Marketing Copy
TEST UNSUCCESSFUL! — Likelihood of AI Content: 43.7%
Undetectable AI Test #3: Cover Letter
TEST SUCCESSFUL! — Likelihood of AI Content: 80.5%
HideMyAI Test #1: Essay
TEST SUCCESSFUL! — Likelihood of AI Content: 67.9%
HideMyAI Test #2: Marketing Copy
TEST SUCCESSFUL! — Likelihood of AI Content: 66.7%
HideMyAI Test #3: Cover Letter
TEST SUCCESSFUL! — Likelihood of AI Content: 83.6%
Overall Tally
Overall, even with AI bypass tools, Copyleaks was able to identify that 9 out of the 10 texts were AI-generated. The one that managed to squeak by Copyleaks’ detection was only able to do so with a score of 56.3% likelihood of being human-written.
TEST CASES | COPYLEAKS AI LIKELIHOOD SCORE |
ChatGPT — Essay | |
ChatGPT — Blog Post | |
ChatGPT — Cover Letter | |
ChatGPT — Prose | |
Undetectable AI — Essay | |
Undetectable AI — Marketing Copy | |
Undetectable AI — Cover Letter | |
HideMyAI — Essay | |
HideMyAI — Marketing Copy | |
HideMyAI — Cover Letter | |
AVERAGE |
Can You Trick Copyleaks?
Is it possible to trick AI detection tools?
That's what I sought to find out in this section. And the answer? Not exactly.
To elaborate on that, I must first tell you how I tried to trick Copyleaks. I started by generating a short essay about global warming using ChatGPT. Here’s what I got:
Next, I tweaked the text using the following methods:
- Be concise
- Avoid using jargons
- Remove unnecessary fluff
- Limit the supporting examples and repetition
Once Copyleaks scanned the modified content, I was not surprised that it was 56.8% likely to be human-written. And this is the reason why…
Remember when I said “Not exactly” earlier? I’ve tweaked the original AI-generated text so much that, at a certain point, I’d say the text ceased to be ChatGPT and instead became…well, me.
So, yes — you can trick Copyleaks. But, is it really a trick if the text was almost entirely rewritten by you?
Pros & Cons of Copyleaks
PROS | CONS |
|
|
How Much Does Copyleaks Cost?
The basic feature of Copyleaks’ AI detection tool is completely free. However, if you’re interested in a more robust AI detector and a plagiarism checker, Copyleaks’ premium version depends on how many pages you’re going to scan per month. It starts at $10.99 for 100 monthly scanned pages ($9.16 monthly for annual billing) and scales up to $679.99 for 10,000 pages ($679.99 monthly for annual billing).
As for the Gen AI Governance and AI Grader tools, the price is subject to discussion when you book a demo with Copyleaks.
The Bottom Line
No AI detection tool can guarantee 100% accuracy — but Copyleaks is pretty darn close.
I’ll admit, I was initially skeptical about their 99% accuracy claim. However, after testing the tool myself, I’m happy to report that the expectations they set were met. Not only was it able to detect ChatGPT content every single time, Copyleaks was also able to flag several AI-generated texts modified by bypass tools.
But, no matter how good a detector is, false positives and false negatives are inevitable. This is something you need to keep a close eye on, especially if you’re an educator, as it could derail a student’s life. As I always say: exercise caution. Nothing beats human intuition.
To sum it all up, Copyleaks is definitely worth a try. Once you've tested it, please do tell us your experience in the comments below.
Want to Learn Even More?
If you enjoyed this article, subscribe to our free newsletter where we share tips & tricks on how to use tech & AI to grow and optimize your business, career, and life.