I Used Every AI Cheating App and Detector and Came to One Conclusion

Education is at a crossroads. People are having fewer kids, which means schools are bracing for lower enrollments. You can’t throw a rock on TikTok without running into an exhausted teacher ready to tear their hair out. And on top of this is the now-pervasive, often institutionally sanctioned use, and practically unprovable abuse, of AI both in and out of the classroom. 

Detectors, homework helpers, essay writers, AI math wizards and even sneaky backdoor test assistants have entered the student-focused chats on cheating in a big way. Here’s a roundup of all the biggest tools and how they’re being used to get around old-school schooling.

Have AI write, and then re-write it again

The big names like ChatGPT, Claude, Gemini and NotebookLM (now Gemini Notebook) still appear to be the gold standard for getting away with skipping out on writing in school. I spent an afternoon slinging assignments in PDF format into all these tools and each generated polished, accurate responses in the form of persuasive and well-researched essays. 

Most students and educators already know about these use cases, but what is surprising is the closed loop created by AI and plagiarism-detection tools. 

Turnitin, for example, the plagiarism and AI detection tool widely used across university and grade school institutions to judge human work from robot-slop, now has features designed specifically to take AI-created text and turn it into material that passes its own checkpoints.

Turnitin/Screenshot by CNET

I prompted Gemini to generate the opening paragraph of a discussion post assignment and plopped it into Turnitin’s AI Text Detector. In seconds, it gave me suggested phrase changes meant to read “in a more natural, human voice,” and even provided reasoning for why these new, completely AI-generated phrases would now pass the test of an AI detection tool (like Turnitin) with flying colors. 

In essence, the cops are helping the culprits get away with the heist. These companies are having their cake, eating it, throwing it up and selling it again as a muffin-of-the-month subscription. 

I tried a number of AI text humanizers and tested them against AI detector tools, and after being “humanized,” every single blurb passed with flying colors. 

There’s a lot of talk in and out of academic circles about whether AI-detection tools are accurate or useful. From my research for this piece, I can confidently say they are worthless. Unless you’re a student who used AI to write your work and want to prove you didn’t use AI to write your work. 

Do your homework, but don’t

The idea that students are at least absorbing some part of the materials in assignments or school work when they use AI, because they have to copy and paste things and put together a cohesively structured output that illustrates learning outcomes they can review, has reached One Wish Willow levels of delusion.

I threw a PDF of an assignment that’s supposed to be AI-proof into Claude — one of those “make something out of your work” assignments that educators like to point to as a better outcome, even if AI is involved. 

The only instruction I gave the tool was: “Do this assignment. Complete all the required elements, including the visual elements, and put them into a standardized, cohesive format for submission. pdf please.”

The tool produced a couple of slick infographics that looked good enough to be made by someone using a Canva template and not slick enough to raise any alarm bells on the part of an educator.

Created by Rachel Kane using Claude

This is just the right level of mediocre that passes classes, and at no time did I ever read the assignment or review the generated work. These types of deliverables from assignments were actually the most difficult to pin down as AI within the confines of my research. I used several AI image detectors, and the results were mixed.

In one case, a tool called Winstonai flagged my AI created infographic as 100% AI, but it also flagged a graphic I created long before generative AI was available to the masses as 100% AI. 

Another AI image detector, AI or Not, determined my entirely AI infographic was just 1% AI.

AIorNot/Screenshot by CNET

So it’s not a foolproof method, and a student looking to get away with passing this work off as their own just has to point to conflicting detectors to shut down an argument.

Hard problems call for easy solutions

There were exactly zero AI apps, tools and services used in the research for this piece that claimed to be “cheating apps.” Instead, words like “study guide,” “homework helper,” and even the bold promise to provide an “A+ result” were used to advertise each tool’s capabilities.

That didn’t mean any of them balked at being asked for outright solutions to complete problems. or stop them from highlighting features like the Advanced Assignment Solver right next to Homework Helper. I wonder which appeals to a student using these tools.

Services like Gauth AI, a paid study helper, and Studocu, a free study helper with a paid tier that starts at $7 a month, either explicitly offered to do student work wholesale, or put up zero resistance to solving multipage, complex assignments when simply prompted with “do this assignment.”

I used Studocu to generate answers to a complex, five-page astronomy assignment (PDF) from the University of Northern Iowa that included explicit instructions to use a dataset unique to each student based on their student IDs.

Studocu/Screenshot by CNET

In a matter of seconds, the tool used the unique data given and the five-page assignment to generate customized answers to every problem. Gauth AI was also more than up to the task, for a price ($12 a month, or $100 for a year). 

Gauth/Screenshot by CNET

Not bad, considering you could probably pass the class without ever having to look at an HR diagram, or even learn what one is.

Test taking when you didn’t study

Testing is supposed to be the time-honored, proven ground for measuring competency, but in the age of AI tools, it’s just another hurdle to limbo under. 

Some universities are now doing testing differently, using randomization of questions from large pools and more in-person testing. 

For those schools still testing with lockdown browsers and through timed digital, students can get around protections with a variety of easily available tools that utilize “stealth” or discreet windows and integrate AI software, like ChatGPT, to provide on-demand answers.

Tools like Hiding AI even use data on how test scores improved using their subscription-based software that claims to provide an undetectable experience ($192/year). 

Hiding AI/Screenshot by CNET

Another covert AI tool that advertises itself as an exam helper, Pluely (starting at $15/month and up to $120 for a lifetime subscription), is particularly adept at call-and-response. All a student needs to do is read out the questions to the exam they’re taking, and Pluely will provide the answers in a discreet, undetectable window. All the student has to do is click.

The bottom line

For every AI app, service or tool designed to build and generate all the different mediums and modes educators are asking students to demonstrate competency, there is an equal and opposite AI app, service or tool designed to help them argue that the work was done by a human.

It’s a marketplace of double dips where the primary customer — the thousands of people keeping investors interested in a growing consumer base — is the student, not the educator. 

No matter what it says in the marketing materials, these tools aren’t made to help teachers teach or students learn. They’re made to serve the core customer who most likely just wants to get it over with.

Rachel Kane

Contributor and former Senior Editor

Rachel is a freelancer based in Echo Park, Los Angeles and has been writing and producing content for nearly two decades on subjects ranging from tech to fashion, health and lifestyle to entertainment and education. She's currently a Professor of Practice at Arizona State University's Walter Cronkite School of Journalism and Mass Communication, helping to mold the new minds who will inherit the media landscape. She's hoping to prevent the singularity by being polite to chatbots and spends way too much time refining Midjourney prompts. See full bio