Gotcha! What’s the Best Way to Catch AI-Generated Copy?

Wondering if all that smarty-pants copy your writer/editor just sent you really was the polished piece of brilliance resulting from years of study, hours of research, and untold effort in the form of carefully crafted word-smithing??

OR…

Was all that lush and lilting language, only the flippant and frothy phlegm of some cold-hearted, coughing AI engine, offering only the dull glint of metallic malaise where once you saw the sparkling effervescence of a human soul??

The accomplishments of modern-day computational acumen in the form of AI have now presented us with no less than 24 AI detection services with names ranging from the straightforward (Writer) to the scary (Quillbot) to the oddly maternal (Hugging Face). They all call out to us, but do they work?

As noted in this Forbes article written by Jodie Cook:

“A copywriter ran the Declaration of Independence through an AI content detector. The result? It's 98.51% AI-generated, despite being written in 1776.”?

Damn. The document that inspired the birth of a nation was somehow AI-generated over 200 years ago. Possibly by aliens.?

According to AI itself, in the form of some smart AI detector, the Bible was also AI-generated. These scandals will keep coming, my people. Hold on tight, Shakespeare aficionados.?

Dianna Mason is the content specialist whose research uncovered the Declaration of Independence Scandal. And she says, “AI content detectors don’t work.”

First of all, how DARE she! Secondly, yes, that’s true.?

Fact: People don't do things for free unless they’re super-bored and/or obsessed with that thing. And companies, since they MUST make money, don’t do ANYTHING for free. Therefore, it’s no surprise that these AI detectors also offer paid “humanization” services–also carried out non-ironically through AI–to supposedly save your content that they flag as “AI-generated.”

What does their “humanization” look like? Hilariously horrible.?

Here’s an example by one of the many AI detectors/humanizers out there. This AI detector is called “Duey.” (You’ll see why I might call it “Don't-y”):?

Before: Ever wished you could teleport to a meeting halfway across the world? VR can make that a reality.?

After: Ever before desired you can teleport to a boardroom halfway throughout the globe? Virtual Reality can make that a truth.?

Does anyone ever really want to go “halfway throughout the globe?” Which globe is that? The globe on your desk or an ESL teacher’s globe in Bangladesh? And how does one “make” a “truth?” Anyone else here fondly recalling Derek Zoolander’s Center for Kids Who Can’t Read Good??

Why is this happening? Simply put, AI uses algorithmic methods of thesaurus-izing (I just made that word up) sentences to strangle the life out of them. In the process, it makes your copy sound worse than awkward. It renders entire sentences meaningless.?

Here’s an interesting question I asked ChatGPT:

“Doesn't an AI platform that promises to analyze whether content is AI-created while also offering to "humanize" for a price, present a conflict of interest?”

ChatGPT answered:

“Yes, it does present a conflict of interest. If the AI platform is offering to ‘humanize’ content for a price, it may have an incentive to inaccurately label content as generated by AI in order to upsell its humanization services. This could compromise the integrity and accuracy of the platform's analysis of the original text.”

So even AI admits that its human Marketing Masters offering these AI detection services may have ulterior motives for gaily and generously flagging content as “AI-generated.” Of course, that motive would be money.?

AI detectors admit they are just guessing when they tell you content has a certain percentage chance of being AI-generated. You statistics geniuses (and card-counting poker players) will get a kick out of this; you understand that the “law of averages” is not actual legislation.

Percentages are funny that way. If you're not careful, people (or AI, guided by certain people) can quite easily twist your perception of truth by throwing a percentage at you at a high enough velocity–say 96.2%–and watching you struggle as it hits you, your face frozen, save for your eyes, shifting sporadically from left to right, desperately in search of meaning.?

For instance, did you know that 49% of all doctors graduated in the lower half of their class? Sounds positively shocking, right? But it’s true.?

Another writer/editor named Dana Yewbank ironically posted a list of over 100 single words and phrases she calls “The official comprehensive list of tell-tale signs someone used AI in their writing!” As Dana astutely pointed out, words/phrases like…

• "significant"
• "characterized by..."
• "encompass"
• "conceptualize"
• "amplify"
• "emphasize"?

…are simply slightly cheesy words, that ordinary people blithely use every day. People. They are not words that AI has somehow magically commandeered to use solely for its own evil purposes.?

Let’s think about this for a moment: How does AI generate content anyway? It scrapes the interwebs for info faster than a college kid on trucker supplements before finals.?

And what does it read? Other human and AI-generated content. It collects these, and sometimes “hallucinates” or lies when it can’t find what it’s looking for. (More on that in another exploration, since that stuff is ripe with hilarity.)? Then it spits this out as requested content.?

AI is trained on new content continuously. The results change, evolve, and sometimes devolve (another topic for discussion under the concept of model collapse, which happens as AI trains more and more on itself) according to these sources of data. This data is constantly in flux, ostensibly reflecting a wider array of materials to draw from every moment.?

This is why there cannot be a set list of “Gotcha” AI words to look for in content. AI content is changing all the time.?

So if you can’t rely on some fancy (expensive or “highly recommended”) AI detector/humanizer to ensure your content is top-notch, what CAN you do??

There is one fool-proof, cutting-edge technique. Consider this carefully.?

You can read it yourself. You can decide if you like it. You can share the piece with another human you respect and trust. You can shout it from the rooftops, or the hills (which may or may not be alive), like Maria in “The Sound of Music.”?

And if this piece resonates with you, if you feel that unmistakable, ineffable shiver of knowing from the tips of your toes up to the top of your head, you are in the presence of Truth.?

How do you solve a problem like Maria? How do you hold a moonbeam in your hand?*

*?Lyric from the song, “Maria” by Oscar Hammerstein II and Richard Rodgers, two amazing humans.