With the release of ChatGPT a few months ago, anybody is now able to easily create papers, essays, projects, and business plans by typing a few words and clicking a button.
There are plenty of ways to "check" if something has been written with AI but the issue is these tools don't really determine anything provable. It's a double edged sword to flag something as AI-generated, when these predictors are robots trying to predict things themselves. Like seriously – the way AI detection works is by reverse engineering AI writing – its an extreme form of predicting things.
Regardless of your use for ChatGPT, until the world figures itself out, you may find yourself needing to bypass AI detection to ensure your content isn't erroneously flagged or dismissed. With the increasing reliance on AI-written content, there has been a rise in AI detection tools, which, as we discussed earlier, aren't always accurate.
So, how can you effectively use ChatGPT to generate content without getting caught by these sometimes overzealous algorithms? The answer lies in understanding the limitations of AI detection and adapting your writing approach accordingly. The easy solution is to use a word scrambler like Undetectable AI, the harder solution is just to write higher-level writing. Let's get into it!
How to Pass AI Detection for ChatGPT Writing
It's such a conflicting scenario, students are getting flagged as cheaters when they don't even have a ChatGPT account! And then you have students that miraculously get through the hoops when they've "obviously" been using ChatGPT to write.
It expands well beyond education. Sometimes you just want to safeguard yourself against getting flagged no matter what your reason may be.
The easiest way to try to fool (or safeguard) a detector is to scramble the words in your writing. Since AI detection works by trying to recreate the text you wrote with AI and see if it's something a computer would write (1 to 1), the more variability in your text means the less likely a computer wrote it.
I'm sure this conversation won't even be an issue in a year from now, as AI will have progressed past the level of any types of detection. The tools just aren't concrete enough and the software is rapidly outpacing the speed of "detection."
Using Undetectable.ai
I've been messing around with Undetectable AI for quite a bit now and have noticed it to be a good scrambler of content. Just keep in mind when using it for more sophisticated writing, as the synonyms and sentence syntax structures it tries to rearrange are harder, resulting in some funky results. But it's a great tool to quickly shuffle things up.
To use the tool, simply paste in your writing & click the humanize button. You should get your new results back in a few seconds. You could also check if what you've pasted is being detected across a few different detectors.
So AI can't create "undetectable AI" since it's using itself to write it! Trust me, no matter what you ask it to do or how creative you ask for a result, it's still going to be based on some predictable writing it was trained upon. See what I mean?
After humanizing the content, you should see the new result in your documents tab. You might have to sign up for an account (but you get a decent amount of free words to scramble). Here's what the new result will look like:
It says it passes every single detection tool, how awesome! I threw it into Content at Scale personally just to make sure and it looks like we're really good!
Use HideMyAI
Another incredible tool like Undetectable (that even offers more customization options) is HideMyAI. This tool offers 550 words for free (with an account) that you could use to bypass ChatGPT with.
I'll ask ChatGPT for another paragraph about something (and I'll use GPT-4 so it's the most advanced):
When I put this into Originality it shows as blatant ChatGPT:
Now let's put this into HideMyAI & change a few settings. I want this one to be very casual, I'll match the perspective to the content I put in, and we'll use the ChatGPT generator:
Here's what we got. I love it. It's super casual & describes everything we got from the ChatGPT result:
How does it do against the strictest AI detector currently available to the public? It's incredible. Not even a single chance it was detected with AI. You should totally try out HideMyAI and mess around with the tone & voice options if you want to rearrange and bypass ChatGPT detection.
Using QuillBot
QuillBot works similar to Undetectable & HideMyAI but lets you click and regenerate the sentences that it scrambled for you. You still get options to change the tone and level of vocabulary, but it's in a slightly different format than before.
I'd say QuillBot takes a bit more work to get perfect but gives you greater customization in a faster and simpler UI to rearrange your writing.
Rearrange Words Manually
If you break down the actual vocabulary that was produced, you'll see it's not very complex. It uses a lot of basic sentence structures even though the story itself is creative. I've highlighted the super simple sentences that a human would be able to divvy up in bold:
As I sit by the window, watching the raindrops race down the glass pane, I can't help but feel a sense of peace. There's something comforting about the sound of rain, the way it pitter-patters against the rooftop, the way it dances on the leaves of the trees.
In this moment, I feel grateful for the simple things in life. The warmth of a cozy blanket, the aroma of freshly brewed coffee, the companionship of a loyal pet. It's easy to get caught up in the hustle and bustle of everyday life, but moments like these remind me to slow down and appreciate the beauty that surrounds me.
I take a sip of my coffee and let out a contented sigh. Life may not be perfect, but in this moment, it's pretty darn good.There's no specific advice you can follow for all pieces of writing, but you get the point! Sophistication is something humans are really good at, and at this current point, robots aren't.
Final Thoughts
So that's it! The key to making AI text from ChatGPT more human-like is to... make it more creative. These robots don't posses the creativity and adaptability to adjust their writing styles based on the level of intelligence a human provides.
Whatever your reason may be, use these strategies to confuse these faulty detectors and stay in the clear. Once again, I really don't think we'll be having these detection conversations a year from now, but it's definitely an important concept to cover while the world is going crazy.
What are your thoughts on ChatGPT and AI detection? Have you seen detectors falsely flag your content in either a positive or negative way? Drop a comment below and let me know your thoughts
