(355) Report: Web Traffic to AI Humanizers and Spinners is Up 518%
Plus, the VICTVS Exam Hack is back. Plus, Abertay University says what it meant by "advanced approach to detection."
Issue 355
Subscribe below to join 4,620 (+14) other smart people who get “The Cheat Sheet.” New Issues every Tuesday and Thursday.
The Cheat Sheet is free. Although, patronage through paid subscriptions is what makes this newsletter possible. Individual subscriptions start at $8 a month ($80 annual), and institutional or corporate subscriptions are $250 a year. You can also support The Cheat Sheet by giving through Patreon.
Report: Web Traffic to Cheating Providers and Essay Mills is Down, Traffic to AI Writing and Washing is Way Up
The outstanding “This Isn’t Fine” substack newsletter put out an important report a week or so ago tracking Internet traffic to cheating providers, by area.
If you don’t subscribe, you should. “This Isn’t Fine” is by Joseph Thibault and Sam Silverman, the former being our Academic Integrity Person of the Year for 2024 (see Issue NY 25). Paid subscribers over at TIF get access to the really valuable Essay Mill Database, which:
is a massive spreadsheet with 438 Essay Mills, Humanizers, and AI Writing Sites along with meta data about each of the sites: location, traffic, URLs, policies, and a bunch more.
As of today, there are >13,000 datapoints, up to date through February ‘25.
But to the report, traffic to traditional essay mills is down steadily, but:
the rate of change is trending to a new normal as the industry pivots as a response to AI writing solutions and AI homework helping sites.
December ‘24 Essay Mill traffic was off -12.97% compared to December ‘23
I’m not sure if that’s good news or not. But what’s definitely not good news is that this cheating traffic has not vanished, it’s just moved. From the report:
we’ve cast as specific a net as we can to highlight tools that are explicitly focused on writing generically enough that use cases include writing essays and academic papers.
These sites are up 191.25% year over year (82 sites).
According to the report, this specifically does not include generative AI sites, but does include humanizer or spinner sites that paraphrase writing to avoid AI detection, as well as essay writing sites specifically. Here, not only is web traffic increasing:
This segment, all told, is about 7x the traffic to Essay Mills and on an upward trajectory.
As important as that is to know, this part is more so:
As a subset, these text spinners and paraphrasers are created for the express purpose of making AI generated text undetectable through AI detectors like Copyleaks, Pangram, Originality AI, or Turnitin (in fact, most score themselves against a growing list of detectors as part of their marketing).
These sites are up 518.05% year over year as a category (16 sites).
It’s true, that spinners and paraphrasers directly sell their functionality as a bypass tool for AI detection — a way to smudge away the tell-tale fingerprints of AI. To me, these sites are more than insidious, worse than the AI engines, for which there may be any number of legitimate, even educational uses. These scrubbers and humanizers exist primarily, even exclusively, to help users commit fraud, to deceive.
Whether they can is an open question. Some AI detectors such as Turnitin and Pangram claim to reliably catch many of the better-known spinning sites.
But whatever — traffic to these fingerprint smudgers is up more than 500%. So, that’s great.
The report also tracked traffic to the “Big Four” sites that sell “homework help.” I refer to them more often as cheating sites — places you can buy answers to homework, or tests, or whatever. From the report:
The mantel of “homework help” has been shifted recently to AI-powered apps. There are quite a few that shoulder that terminology already while pushing social proof (University logos).
The big four are Chegg, Quizlet, Coursehero, and Quillbot.
Year over year they are down -39.71%.
There they are. My friends. Reminder, Coursehero and Quillbot are the same company — a Potemkin Village-level whitewash now known as “Learneo.”
In any case, traffic to these poison peddlers is down. As has been well chronicled, AI is picking their pockets, as students are no longer eager to pay for what they can now get for free.
The report also has some insightful scratch calculations on the size of these various cheating industries:
We estimated the global English-language market to be $4bn in late 2022.
If we assume that traffic is correlated with revenue, Essay Mills are off at least 50% since ChatGPT: $2bn market.
Applying the same estimate to AI site (7x the traffic of our EM collection) that would be $14bn (too high IMO).
However, AI output per page costs about 90% less than an Essay Mill; which would peg AI-enabled cheating sites at about $1.4bn.
To put that in context, the plagiarism detection market is estimated at less than $2bn in 2024
I don’t have much to say about this math, except that it ought to feel daunting. It is. The demand for easy, work-free solutions and credit for nothing is relentless. And too many unethical people are willing to sell those services. Meanwhile, the required investment to deal with it is lacking.
When the house falls down, no one should be able to say they are surprised. Same old story — plenty of people in this universe can act, they choose not to.
Abertay University Comments on its “Advanced Approach to Detection of Misuse”
In Issue 350, we shared news coverage of surging cases of academic misconduct in Scotland. In the coverage, Abertay University reported a significant uptick in cases related to AI use and a high “responsible” rate of 97%:
Abertay University introduced “unacceptable AI use” as a misconduct category in 2023 and last year dealt with 351 cases, with the complaint being upheld in 342 of them.
The news article also refers to the school’s “advanced approach to detection of misuse," as a factor in their increase in cases and high positive determination rates. Being unsure what “advanced approach” means, I asked the school.
Here is the response from a spokesperson:
My understanding on this is that we pay for an advanced version of TurnItIn capable of better identifying AI-generated work than traditional packages.
I am not sure Turnitin has an upgraded package. Maybe it does.
But the point I’d like to highlight here is that when a school decides to look for academic misconduct, to actually spend money to spot it and stop it, they find it. Imagine that.
That reality calls into deep question — once again — what on Earth schools such as Vanderbilt and University of Texas are doing when they simply unplug their AI detection. I mean, I know what they’re doing. They are closing their eyes and allowing cheating. As their policy. Just so we’re clear.
VICTVS Exam Hack is Back
The VICTVS Exam Hack Challenge is one of my favorite events of the year.
And, it’s back. This year, VICTVS has asked if I’m interested in being a judge. You bet I am. Not sure what that means yet, but I will let you know.
In the meantime, this description of Exam Hack is provided by VICTVS, an assessment security provider based in the UK:
The VICTVS Exam Hack 2025 is a unique challenge for students, academics, and conniving minds who want to test their skills in the art of ethical deception. As an exam candidate, competitors (acting alone or with a partner) will take on the challenge of sitting a remote exam and finding a way to outsmart VICTVS's exam software and expert invigilators. The day’s most inventive and creative hack wins a cash prize of $1,000 (USD). You can find more information here: https://examhack.io/
VICTVS hosts this event to stay on top of new cheating methods, ensuring that they continue to adapt and lead the way in exam security and integrity.