With so many AI checkers on the market, it can get overwhelming to sift through them all and see what works best. ChatGPT and other AI bots have changed the landscape of online content creation, forcing many editors and site runners to find a viable AI detection tool.
Although AI can be a very useful tool when it comes to writing content, we have started to see many writers turn in completely AI-generated work, a big no-no that is no different than plagiarism.
Undetectable.AI is a service that can check if an article, essay, or any piece of writing was generated by a human or a bot, and even claims to be able to rewrite them for you to avoid AI flags.
We tested 3 ChatGPT generated articles and 3 articles that were written by a human to see if this tool could pass the test.
What AI Tools Can Undetecable.AI Flag?
Undetectable.AI can currently flag AI-generated content written by GPTZero, OpenAI, Writer, CrossPlag, CopyLeaks, Sapling, Content At Scale, and ZeroGPT. The app offers a wide net that can flag most of the major AI tools currently available.
It can help determine if the work is 100% written by AI, 50%, or 0%, and gives you a color indicator for each AI service analyzed. Although it would be nice to have results a bit more granular, the app does provide reliable high-level information that should be accurate enough to start a conversation with your writer.
Our Methodology
Our testing process was simple – we took three AI-generated articles and three human-generated articles and ran them through the Undetectable AI check software.
ChatGPT generated three 1,000-word articles on the following topic subjects: Healthy Food, Basketball, and World History. Our human writers wrote their articles on the following topics: Traditional Archery, Productivity Tools, and “Frozen” Characters.
Does Undetectable.AI Work?
In our limited testing, Undetectable’s AI checker works exactly as advertised. We tested three 1,000-word AI-generated articles and all were flagged as AI-generated content. The quick and simple UI (pictured below) shows a snapshot of what Undetectable.AI thinks about the analyzed writing.
Healthy Food Results (AI #1):
Basketball Results (AI #2):
World History Results (AI #3):
Conversely, we also ran three articles that we wrote ourselves, to see if the app may slip up and identify genuine human content as AI written. Undetectable works well here, although we would have liked it to be a bit more certain.
Traditional Archery Results (HUMAN #1):
Productivity Results (HUMAN #2):
Frozen Characters Results (HUMAN #3):
As we can see, the app was correct 100% of the time. Even though Undetectable appropriately labeled each example, it is still worth mentioning that it did not pass or fail the checks across the board.
Take a look at the AI #3 example – Undetectable flags the article as AI, but then mentions that it is 50% likely that the writer used GPTZero, Writer, CrossPlag, Sapling, or Content At Scale. It also points out that it completely bypassed the OpenAI check, and only flags 100% AI with Copy Leaks and ZeroGPT.
This is concerning because this article was written using OpenAI’s ChatGPT, something that Undetectable completely missed. Although you can never trust a service like this to work 100% of the time, it is definitely something that you should be looking out for while conducting your own tests.
AI Text Humanizer
Apart from the AI detection tool, Undetectable.AI also has an AI text humanizer. In theory, this tool will take AI-written work and rewrite it to sound like a human.
The app claims that it can bypass the most advanced AI detectors on the market, although the results may vary. It gives you the option to write at a high school, university, doctorate, marketing, or journalistic level.
This humanizer is a large part of Undectable’s offering and commands a pricey subscription fee. You can save a lot of money if you purchase their yearly offering.
We decided to take our three AI-generated articles and run them through Undetectable’s “text humanizer.” After we received our humanized content, we ran them through some AI-content detectors to see if the app generated reliable “human” content.
There are various settings for Undetectable Humanizer, and we decided to have the article be run at a high school reading level, under the “general writing” category.
Our results are listed in the table below (the percentage listed is the likelihood that the content was written by AI):
Article | Originality.ai Score | GPTZero Score | CopyLeaks Score | Writer.com |
|---|---|---|---|---|
AI #1 | 100% | 65% | 90.1% | 5% |
Undetectable Humanizer #1 | 1% | 1% | 32.6% | 0% |
AI #2 | 100% | 91% | 90.3% | 83% |
Undetectable Humanizer #2 | 9% | 11% | 37.6% | 0% |
AI #3 | 100% | 94% | 88.8% | 14% |
Undetectable Humanizer #3 | 2% | 17% | 42% | 0% |
As we can see, the humanized text did fool our AI-content checkers. Even though this looks good at a high level, we started to realize that the “humanized” content was just “dumbed-down” in order to sound more like a human. But there is just one problem, most humans, even when writing at their worst, sound much better than Undetectable’s output.
Take a look below, at Undetectable’s first two sentences from the “humanized” Healthy Foods article:
“In todays paced society maintaining a healthy diet plays a crucial role in promoting overall well being. A balanced and nourishing eating plan not helps control weight but mitigates the risk of chronic ailments boosts energy levels and uplifts mood.”
The human content delivers writing that would never be accepted by an editor, a teacher, or anyone else reviewing your work. So unless you are prepared to do some heavy editing yourself before turning in the work, we would advise against using Undetectable’s text humanizer.
Yes, the output may beat an AI content detector, but it sacrifices the flow, grammar, and readability of the article.
The Verdict
Undetectable is a serviceable AI detector that worked just fine in our limited testing. It is great that the AI detection service is free, giving many editors and writers a viable, cheap option to check on their work.
Even though the service may flag human content as “50% human” in spots, a good editor can usually sift through the article and determine if AI was actually used. We have to remember that these AI checkers are not foolproof and false positives may come through.
One thing we love about Undetectable is its 15,000-character limit for free AI checks. This is far above the norm for many competing sites, and it makes the process very convenient for those who are not looking to sign up for a monthly subscription.
We don’t recommend the service’s text humanizer, as the output only tricks AI checkers with improper grammar and awkward writing. You will probably put in more work editing and rewriting the article than if you just started from the beginning yourself.
If you are on a budget and need a tool to quickly check on an article, Undetectable.AI should work for you. If you are looking for a more advanced “humanizer” service, be prepared to pay a hefty price for an underperforming product.
I’m currently wearing many hats but my primary one is being the Business Manager at BrandBuilders.
My professional journey spans from being a tech enthusiast to evolving into a seasoned serial entrepreneur, with a primary expertise in SEO and Digital Marketing.
I am currently delving into the frontier of Ai, exploring its applications and implications in the ever-evolving digital landscape.
