If you’ve ever used an AI content generator like Claude (for writing) or Midjourney (for art), then you’re probably familiar with how AI content feels. Heck, even if you encounter AI content on a regular basis, you eventually learn how to spot it a mile away.
There’s a certain look to AI-generated art that tells you it was done by a machine. Likewise, there’s a certain quality to AI-written text that – more often than not – gives away its artificial origins.
And if you aren’t totally sure, you can always use AI content detectors to confirm (or debunk) your suspicions.
But, as per usual, technology is catching up.
If there are programs that can detect AI, there are also programs that can fool AI. These programs modify AI-generated content so that they slip under the radar and appear, for all intents and purposes, like they were written by humans.
Exciting? Sure. Concerning? Maybe. But remember that the tech itself is morally neutral. What matters is how it’s used. And as long as it’s used responsibly and ethically, there should be no qualms about allowing such programs to grow and improve.
In this article, we’re going to focus on Undetectable.ai – an AI detection and AI modifier tool that’s currently considered one of the best in the industry.
Testing the Check AI Feature
Currently rated as the #1 Best AI Detector by Forbes, Undetectable.ai stands out as a program that both detects AI and bypasses AI. Want to know if a piece of content is AI-generated? Run it through Undetectable.ai.
Want to make that same piece of content bypass detection?
Run it through Undetectable.ai.
Remember that AI detection tools work by analyzing patterns. And while it’s far from being a hundred percent accurate, the technology is there. There’s a range to just how blatantly AI a piece of content can be, and tools like Undetectable.ai are definitely clued into that range.
How to Use the Humanize Tool Responsibly
Before I show you Undetectable.ai’s Humanize tool in action, we need to first cover what constitutes as responsible use of such technology. Because this technology, though fresh and experimental, is powerful.
And remember Uncle Ben’s (of Spider-Man fame) oft-repeated line?
“With great power comes great responsibility.”
I know, I know; so utterly cheesy. But it’s also utterly true. Just because you can humanize AI-written content so that it slips under the radar doesn’t mean you should.
Without plenty of care and consideration, anyway.
Publishing AI content and allowing people to believe that an actual thinking human being wrote it raises plenty of ethical concerns:
Is the AI content considered intellectual property? If so; whose? And does this eliminate the need for human intervention–further discouraging professionals from using AI to supplement their craft?
Technology this powerful can be used to manipulate, mislead, and misinform.
Imagine paying someone to craft original content and, in turn, receiving something that they had a machine generate. For free. It’s underhanded and (and I don’t think anyone would disagree here) immoral.
So here’s how to use Undetectable.ai’s Humanize Tool responsibly:
- Transparency. If possible, inform your consumers that the content was generated with AI assistance.
- Regulate & Review. Always fact-check the information generated by the AI before publishing it. If possible, get a second or third (professional) opinion.
- Supplement, Not Replace. Use Undetectable.ai’s Humanize Tool to help you understand how to write or rewrite content pieces. Don’t publish what Undetectable.ai writes and pass it off as human.
- Humanize the Content Yourself. After you use the Humanize tool, further modify and/or rewrite the resulting text. This solution may take longer, but it also adds a truly human touch to the generated content.
- Use Your Own Content as Reference. Undetectable.ai has a very clear stance regarding academic dishonesty and cheating: they don’t like it. In this regard, if you’re going to publish Humanized AI content, feed the AI generator your own work (instead of others) as a reference.
Moral of the story? Just don’t professionally and publicly claim Humanized AI Content as actual human content. And especially don’t do it for personal gain at the expense of others.
The Authorship Replication Tool
Aside from the Humanize Tool, Undetectable.ai also has the Authorship Replication Tool. This tool is a powerful piece of AI tech that, as the name suggests, can replicate an author’s writing style, tone, and voice.
Imagine being able to generate a hundred pieces of content that perfectly match your writing style and voice–in seconds. It’s a truly innovative program that any business could benefit from.
But, as with the platform’s Humanize tool, Undetectable.ai’s Authorship Replication tool comes with plenty of caveats for responsible use. At the bottom of their Authorship Replication Tool Guide, they have a section dedicated to the ethical utilization of said tech.
The platform itself states that it doesn’t approve of anyone using its services to “bypass educational AI detection systems.” They subsequently encourage their users to use their platform responsibly in the interest of “maintaining trust and integrity in the industry.”
What does this mean?
By all means, use their Authorship Replication tool to generate content that matches the tone, style, vocabulary usage, sentence structure, and so on of a certain writer. But only do so for professional marketing, branding, and/or advertising.
Here’s a good example: Instagram captions.
Write a couple of Instagram captions the way you would normally write them. Then feed what you’ve written to Undetectable.ai and ask it to write new captions that mimic the style, tone, and voice of the text you’ve given it. This way, the content generated technically counts as yours. You’re the only writer the platform copied!
What would be an irresponsible use of Undetectable.ai’s Authorship Replication tool?
- Asking it to write an academic paper mimicking the formatting and writing style of the smartest kid in the class.
- Having it write an article replicating the tone of voice of an established influencer, and then circulating that article and claiming the influencer wrote it.
- Feeding it someone else’s social media captions, having it generate dozens of new captions that copy the original (voice, style, wit, etc.), and then using those copies for personal profit.
The Don’ts of Undetectable.ai
The ethics of using AI tools remains a hot debate.
There’s no denying that Artificial Intelligence is a useful and powerful piece of tech that so many people in different industries can benefit from–as long as it’s used responsibly.
Undetectable.ai is no different. It’s a tool meant to help, not harm. It’s meant to supplement, not replace. And although its services aren’t perfect, they can only improve over time. We must learn how to use these services without infringing, violating, or disregarding anyone’s rights.
So, in summary, you cannot and should not use Undetectable.ai to:
- Pass off purely AI-written content as written by a human.
- Example: academic papers, research, informative essays meant for public circulation.
- Provide “proof” that your AI-written content is human-written (when it actually isn’t).
- Generate content that spams, misleads, misinforms, or offends any audience (intended or not)
- Impersonate another person with the intent to scam, trick, mislead, or misinform their audience
- Example: using the Authorship Replication tool to mimic the writing style of a known author/writer in order to manipulate their established audience
- Impersonate another person with the intent to defame them
- Personally benefit from Undetectable.ai’s services at the detriment and/or loss of others
The platform itself has a list of Prohibited Activities, and it’s a lot more detailed and extensive than the one here. I highly encourage you to check it out. You might think your Undetectable.ai usage is ethical but it never hurts to be a hundred percent sure.
Using Undetectable AI
First, I decided to start with the source material. I fed ChatGPT the Introduction of this paper from NCBI and gave it the following prompt:
“Please summarize this Introduction into a 150-word definition and explanation for Tess, the Mental Health Chatbot.”
I then asked Undetectable.ai to Check for AI. I set the readability level to “University” (since we’re using a published paper as our base) and the purpose to “General Writing.”
Here’s what it found.
According to the platform, this AI-generated text appears human. It will also appear as human to almost every AI detection tool except Copyleaks and ZeroGPT. Undetectable.ai’s AI Detection Likelihood feature determines how a piece of content will potentially fare when run through other AI detection tools. This gives you further odds on how well the text bypasses detection tech in general.
For the second round, I decided to run a raw ChatGPT response–no source material, just 100% AI writing from scratch. I gave it a similar prompt (“can you please write a 150-word explanation of Tessa, the Eating Disorder Chatbot that got shut down?”) then ran its response through Undetectable.ai with the same parameters.
Lo and behold:
We got one.
If you notice, the platform’s AI Detection Likelihood also predicts that the raw AI text will register as AI with a majority of other detection tools like Sapling and Content At Scale.
Now that we’ve got a juicy piece of AI content that will actually ping as AI, let’s check out Undetectable.ai’s ability to bypass AI.
Testing Undetectable.ai’s Humanize Tool …
Undetectable.ai has a Humanize tool that basically rewrites AI-generated content to make it seem like it was written by a human.
Pretty self-explanatory, yes, but just so we’re all on the same page.
It does this by deleting words, rephrasing paragraphs, modifying sentence structure, and so on. Here’s what the platform has to say about this feature, too:
“…seamlessly scale your content output without the risk of getting flagged or sounding too robotic. Create content at scale that sounds perfectly human for your blog, social media posts, white papers, presentations, reports, and much more.”
It’s also worth noting that Undetectable.ai’s Humanize feature comes with a level of customization. You can choose to make it More Readable, Balanced, or More Human.
They recommend the More Readable option for blogs and marketing copy.
The Balanced option is the best choice if you’re after quality.
The More Human choice is recommended for “overly aggressive” or inconsistent AI detectors.
You can essentially get a minimum of three “humanized” options with Undetectable.ai. And your choice does change the modified content in significant, noticeable ways.
Using the raw ChatGPT text from the previous section (the one about Tessa AI) as our guinea pig, here’s what the program returned for the More Human option:
Here’s what we got for the Balanced option:
And for the More Readable option:
You may have noticed that the More Human iteration has quite a few grammatical errors; run-on sentences, improper punctuations, awkward sentence structure, etc. The More Readable result, on the other hand, would pass Grammarly with flying colors. And the Balanced copy is, predictably, a mix of the two.
These errors, although off-putting at first, are done on purpose. Text that’s too perfect and/or too structured could read as AI.
To err is human after all.
Regardless, all three “humanized” copies should now – ideally – pass AI detection.
… Versus Other AI Content Detection Tools
Here’s what happened when I ran all three variations through the AI content detection tools Undetectable.ai listed (GPTZero, OpenAI, Writer, Crossplag, Copyleaks, Sapling, ContentAtScale, and ZeroGPT).
More Human: 4/7
- Sapling – 82.7% AI
- GPTZero – 70% Human
- Writer – 100% Human
- Crossplag – 100% Human
- Copyleaks – 100% AI
- ZeroGPT – 100% Human
- ContentAtScale – Reads Like AI (3 human sentences, 3 AI sentences)
Balanced: 3.5/7
- Sapling – 100% AI
- GPTZero – 97% AI
- Writer – 93% Human
- Crossplag – 100% Human
- Copyleaks – 100% AI
- ZeroGPT – 100% Human
- ContentAtScale – Hard to tell (3 human sentences, 3 AI sentences)
More Readable: 3/7
- Sapling – 100% AI
- GPTZero – 97% AI
- Writer – 87% Human
- Crossplag – 100% Human
- Copyleaks – 100% AI
- ZeroGPT – 100% Human
- ContentAtScale – Reads Like AI (no percentages available) (1 sentence as human, the other 5 AI)
Since Undetectable.ai recommends the More Human option for “overly aggressive” detectors, it comes as no surprise that this variation won the most “human” pings–with four out of the seven platforms believing a human being wrote the text.
The Balanced and More Readable versions don’t do too bad, either. I gave Balanced an extra half point since Content At Scale rated it as “hard to tell.” More Readable managed to sneak past Writer, Crossplag, and ZeroGPT, but it seems the other programs caught on.
Still; the fact that this purely AI-written text was transformed into content that could fool more than one AI detection program is impressive. And it did it in a matter of seconds, even!
So the technology, while not perfect, is there.
Undetectable.ai’s services are quite reasonably priced, so they can be considered “accessible.” Anyone with a budget can freely use their Humanizer tool if they wish.
Thus, responsible use is highly encouraged. If you’re going to have AI rewrite AI so that it doesn’t read like AI, only do so when the situation calls for it. Before you click “Humanize,” ask yourself: is the result going to harm anybody?
That’s as good a litmus test as any.
