The Best AI Detection Tools to Catch Cheating and Plagiarism

OpenAI’s ChatGPT, an artificial intelligence, has embedded its way into higher education and can write essays, pass exams, and teach you a good dorm dinner recipe.

Since the tool made national headlines in 2022, professors and universities have had to address it in syllabi or policy, some schools have embraced and recommended students’ use of ChatGPT and now others are developing degrees and colleges for artificial intelligence.

AI-generated content isn’t limited to just text, it’s becoming increasingly common for visual and audio media. Some detectors can check if AI partially or wholly created your photos or audio.

These tools are not foolproof, and the companies that created them do not claim they are.

According to TechCrunch, “As text-generating AI improves, so will the detectors — a never-ending back-and-forth similar to that between cybercriminals and security researchers… That’s all to say that there’s no silver bullet to solve the problems AI-generated text poses. Quite likely, there won’t ever be.”

Still, these tools can at least detect lower-quality AI-generated content and give users a feel of whether or not a text was human-generated. Generally speaking, the longer the body of text in a passage, the better the tool works.

According to Search Engine Journal, OpenAI plans on introducing a cryptographic watermarking feature. By embedding statistical patterns into word choices and even punctuation, the watermark would make it easy for a detection tool to catch it.

Scott Aaronson, a computer scientist working on the OpenAI watermarking project, said, “We want it to be much harder to take a GPT output and pass it off as if it came from a human. This could be helpful for preventing academic plagiarism, obviously, but also, for example, mass generation of propaganda… Or impersonating someone’s writing style in order to incriminate them.”

Until watermarking features are implemented, however, AI detection tools may be needed to help prevent cheating and plagiarism. Here’s a list of the current tools that students and professors can use to help detect if something was written using AI.

OpenAI discontinued its AI Text Classifier last July; it was a work in progress that correctly identified “26% of AI-written text (true positives) as ‘likely AI-written.'” The tool yielded significantly less positive results than other detectors that claim up to 99.6% accuracy.

On April 4, TurnItIn released their AI writing indicator, which will pair with their pre-existing programs, including Turnitin Feedback Studio (TFS), TFS with Originality, Turnitin Originality, Turnitin Similarity, Simcheck, Originality Check, and Originality Check+.

TurnItIn announced AI writing detection is now a feature for its iThenticate product on Nov. 1, 2023, as part of its 2.0 update. iThenticate specializes in academic research writing integrity.

For the most part, the tool is intended for educator-use only. Students will not be able to use the indicator tool from their accounts.

The detection model will show an overall percentage of the document that might have been AI-generated after breaking down the submission into five-to-ten sentence segments, and rating those sections on a scale of 0, human-generated, to 1, AI-generated. BestColleges senior writer and college instructor, Mark Drozdowski, found the tool to be “remarkably accurate.”

Beginning Jan. 2024, only customers with licenses to Originality or TFS with Originality will have access to the detection tool, according to their site.

Edward Tian — a computer science major and journalism minor at Princeton University — recently created GPTZero during his winter break. More than one million people have already used the app since its launch on Jan. 3.

Tian tweeted, “in short, there’s so much chatgpt hype going around. is this and that written by AI? we as humans deserve to know!” His original tweet about the app’s launch received over 7 million views on Twitter.

According to NPR, Tian trained the bot to detect two different factors: “perplexity,” which measures the text’s complexity, and “burstiness,” which compares sentence length and complexity variation.

Tian recently released a model update called GPTZeroX, intended primarily for educators. The updated tool highlights the parts of the text that are AI-generated, supports larger inputs — including files for an entire class — and operates faster than the pilot version, according to Tian’s Twitter. According to Tian, these were the most requested features by educators.

GPTZero announced a partnership with the American Federation of Teachers on Oct. 19, 2023 to bring metrics and writing visualizations for students and the GPTZero Writing Report for students to verify authenticity.

Created specifically for educators and publishers, Winston AI’s detection tool is one of the best for checking if educational writing was written by a student or with the assistance of AI.

It is capable of detecting the upcoming version of ChatGPT, known as GPT-4.

Their Optical Character Recognition (OCR) technology takes out text from scanned documents, including hand-written text. The tool also supports French, and is gearing up to support more languages soon.

The detector lets users know, on a percentage scale from 0-100, the odds that the copy was written by a human. The software also detects plagiarism, and provides a list of duplicate text found on the internet with printable reports available.

Users can try the tool for free for up to 2,000 words, but will have to pay $14 per month for up to 80,000 word scans.

Ten years in the making, Copyleaks’ AI Detection tool has been trained to sniff out content written by other AI chatbots. Their tool is the first of its kind to detect AI across several languages, including English, Spanish, French, Portuguese, German, Italian, Russian, Polish, Romanian, Dutch, Swedish, Czech, and Norwegian.

Copyleaks claims a 99.12% detection rate accuracy, but admits that the accuracy to detect AI-generated creative writing is typically lower than other content styles.

Copyleaks stands out because not only does it provide a browser-based detection, but it easily integrates with some of the most popular learning management systems (LMS) used by educators, including Canvas, Moodle, Brightspace, Blackboard, Schoology, and Sakai.

The company is working to detect AI-generated text that has gone through a text spinner or has been manipulated to avoid detection. However, it comes with a price. For 1,200 credits, users will have to pay $8.33 per month, and the price rises with more credits.

Originality.AI’s detection tool is designed to be used by writers and web publishers. It claims a 94% detection rate on GPT-3 generated content, and it is easily downloadable as an extension on Chrome for reviewing any piece of content a person is writing or reading on the web.

The detection module scores text, and even entire websites, on a scale of 0-100 on the chance that it was AI-generated. The trial for the detection tool is free, but after the trial period ends, it will cost $0.01, or one credit, per 100 words for users.

The tool can also scan for plagiarism. According to Gold Penguin, an agency that writes free tutorials about the latest AI software, “With the abundance of information available online, it’s easy to accidentally plagiarize without even realizing it. Originality can be a valuable tool for students looking to verify the originality and authenticity of their work before submitting it for a class or assignment.”

Content at Scale is best known for its promise of generating full length and SEO-optimized blog posts with just a few clicks, all the while surpassing even the toughest AI detectors. This is why its version of the AI Detector, which measures a text’s human content score, is one of the most popular free detectors on the market.

Users can input up to 25,000 characters into the tool.

Content at Scale also offers an AI image detecting tool.

Just like Originality.AI’s model, Writer’s AI Content Detector is intended to help writers make decisions about what content they are posting online. But anyone can use the tool for free. Users can enter 1,500 characters at a time. After entering text into the box, the tool will spit out a detection score — the higher the percentage, the higher the likelihood the content was written by a bot.

The tool can now detect GPT-2, GPT-3, GPT-3.5, and ChatGPT, according to Gabe.

Scribbr’s AI detector requires no sign up to use. Scribbr recommends text between 25 and its cap of 500 words to get a good result. The site says students can find out whether their sources were created by AI or if their work accidentally contains AI sections.

The tool uses a bar to highlight what percentage of the text was written by humans and what was written by AI.

The AI detector developed by former researchers at University of California Berkeley, Stanford University, Google, and Meta, uses a similar machine learning system to one’s that generate content like ChatGPT. It generates the probability it thinks each word is AI-generated or not. According to Sapling, it is over 97% accurate and recommends longer text inputs to increase accuracy.

Sapling’s free version allows up to 2,000 characters and the pro and enterprise versions allow up to 8,000 characters.

Sapling also offers an internet browser extension.

ZeroGPT’s AI detector supports all languages and highlights each sentence written by AI with a percentage of how much AI is in the text. The tool generates a PDF report for each submission for plagiarism-free proof.

The tool also allows multiple file uploads at once.

ContentDetector.AI is a free online AI detector with unlimited check and word count for text with up to 85% accuracy. The site is aimed towards bloggers with information on how Google may affect a blog’s search results with AI-generated content. The model uses GPT-2 and BERT models to detect plagiarism.

AI or Not’s free plan allows users to input 20 web image checks, 100 application programming interface checks, and five audio checks. AI or Not can also analyze audio and highlight voice and music samples that were created by AI.

The photo AI detector offers multiple plans offering more details on if a photo was created by AI. The free plan gives a single percentage likelihood if a photo was AI generated. The $10 basic plan breaks down the likelihood the photo was created by a human or AI photo generator DALL-E 2, MidJourney, Stable Diffusion, or other.

Illuminarty also offers a browser extension and an AI text analysis feature.

As higher education normalizes AI in the classroom, more companies are making tools to discourage using AI for plagiarism. Coursera is the latest company to develop an AI-driven suite of tools for educators.

Coursera announced an AI academic suite pilot for educators on June 11, 2024, to be released in the coming months. The suite helps educators generate and grade assignments, improve student academic integrity, and enhance learning through a plagiarism detector, AI-generated content, and AI-generated feedback.

The suite will catch plagiarism from prior submissions, have an AI proctor, and block unauthorized websites and resources during exams. Educators can also utilize the “graded item locking” feature to ensure students complete every lesson item, such as readings, videos, and questions, without skipping ahead.

Coursera’s CEO Jeff Maggioncalda told CNBC TV-18 that, unlike other plagiarism detectors, Coursera doesn’t automatically penalize students for cheating. Instead, students fill out a questionnaire given by the AI. Depending on their answers, the bot will flag a student’s answers to an educator if it detects a student may have cheated. From there the educator can determine if the student cheated or not.

The AI proctor will utilize a student’s camera to verify their face to that test and watch to see who or what is in frame to prevent cheating. The AI proctor also won’t penalize or indicate that a student is or isn’t cheating, but will send that notice to the educator where they can watch the video feed.

“Online learning has become a powerful tool for institutions to better prepare students for a rapidly changing world,” Maggioncalda wrote in a press release. “However, universities must ensure it meets the rigorous standards required for academic credit. While Generative AI introduces new risks for student misconduct, it also provides unprecedented opportunities for universities to enhance academic integrity at scale.”