“So, who’s your crush?” my friend asked eagerly.
We sat on the floor of her basement, surrounded by crumbs and an unnecessary amount of pillows. I avoided her question, “I don’t have time for that stuff.” “Oh, tell me! There has to be one boy.” She was going to interrogate me until I gave a satisfactory answer. I sorted through the people in my science class and smiled, “Finn, from your science table.” She grinned, looking very pleased with this answer. “Of course! I knew I saw you sneaking glances at him!” She was wrong, I was glancing at her.
- By Olivia Sharma, 15, a winner in the The New York Times’ 100-Word Personal Narrative Contest
I asked ChatGPT to “write a short story (100 words) about a teen crush. It should have a surprising ending.”
Here is the result:
Lila had always admired Jason from afar, the soccer star with the messy hair and easy smile. She rehearsed conversations in her head, but her words always fumbled when he passed by. One afternoon, he approached her locker, heart pounding, she managed a shy smile.
“Hey, Lila, I need to tell you something,” he said, glancing nervously at his shoes. Her heart leaped.
“I’ve been dying to ask… does your brother like me?”
Lila blinked. “What?”
“Yeah, we sit next to each other in chemistry, and I think he’s really cool,” Jason continued, oblivious to her sinking heart.
Similar story; but the original just hits differently, doesn’t it? Even in only 100 words, the AI version lacks the depth and dimension that create emotion.
There’s a reason we aren’t outsourcing our clients’ content to AI
Several, in fact.
In a previous blog post, we wrote about how generative AI content creation would be business suicide.
To understand why, we need to go back to basics: why do we do content marketing? Generally speaking, the goal of content marketing is to attract, engage, and retain a target audience. How do we achieve that? By creating and distributing valuable, relevant content.
Can’t AI do that? It can, to an extent. But there are plenty of other businesses out there competing for your audience’s attention, and your content needs to stand out in a sea of sameness. Standing out means coming up with something new, unique, and creative—technically speaking, this is something that AI is (currently) unable to do.
And let’s not even get into AI’s issues with accuracy, which can be confusing, offensive, and downright unsafe. Those are well-known reasons to pump the brakes. Here are four reasons we aren’t outsourcing content production to AI.
1. Good content is not a check-box activity
There’s a distinction between check-box content and intentional content. Generative AI content creation may fill a content calendar, but does it attract, engage, and retain a target audience?
Our clients hire us to create authentic, unique content that’s written in their individual voices. Perhaps we could train AI to do some of that for us. But the tech isn’t sitting in our meetings (or is it?), tuning into the nuances of non-verbal communication, and sensing a vibe. It takes a human to notice (whether consciously or subconsciously) these intangibles, to make connections between ideas, and to come up with novel approaches to age-old problems.
AI isn’t here for that.
You may also be interested in: Content aligned with business goals drives results
2. We care about our craft
As creators, we can’t afford to let our creative muscles atrophy. Sure, we can get AI to generate a passable output, but there’s an inherent danger in becoming a bystander to the process. We’ll lose our edge. And, as a team, we prefer to stay sharp.
Along with degrading our skills, using AI for writing also takes away the heavy lifting of learning about a topic enough to explain it to the target audience and make a cogent argument—making us all dumber over time. Conversely (and currently), we get better at creating content for our clients over time because we spend so much time working with their material. By becoming Subject Matter Proficients, we’re able to get a better grip on our clients’ products, services, and industries—knowledge that enhances the marketing strategies we deliver.
3. Regurgitating is for the birds
A chink in the AI armor is the presence of entropy, a complex concept that describes a method for measuring disorder. Entropy is either stagnant or increasing over time, and in AI, it leads to model collapse. Large language models (LLMs) are trained on text data from a variety of sources, such as books, websites, scientific articles, and social media. When that content is AI-generated, LLMs are essentially being trained on data generated by earlier versions of themselves, and they lose touch with the original data distribution.
In an article titled AI Entropy: The Vicious Circle of AI-Generated Content, expert David Sweenor explains, “Model collapse refers to a degenerative process where, over time, AI models lose information about the original content (data) distribution. As AI models are trained on data generated by their predecessors, they begin to ‘forget’ the true underlying data distribution, leading to a narrowing of their generative capabilities.”
I strongly recommend reading Sweenor’s entire article—he has some great examples illustrating the concept and links to academic research backing up his claims.
TL;DR: LLMs become less accurate and more exaggerated over time as they digest their own generated content.
Who wants to read dumb content? Not me.
4. Imitation is the sincerest form of falling flat
Plagiarism has been a dirty word drilled into us from our school days. When you use generative AI content creation that scrapes others’ publicly available ideas from the web and patches them together without your thoughts, analysis, and writing style, this could—at worst—be construed as plagiarism. And even if it’s generating plagiarism by proxy, the end product is still going to be just that—generic.
Creating text that is too similar to existing content could also lead to SEO penalties, as search engines are on the lookout for duplicate content (content that is identical or very similar to content available elsewhere on the web).
And who wants to read platitudes, crutches and cliches like: “In today’s dynamic world”, “intricate tapestries,” “diving or delving into” anything, and “testament to”?
You may also be interested in: 10 SEO trends you shouldn’t ignore in 2024
Are we anti-AI?
No. In marketing (and the world of knowledge work more generally), being a Luddite risks rendering you irrelevant. But there has to be a balance.
Here’s how our content writer, Shea Karssing, describes her use of GenAI: “ChatGPT is like a helpful intern (that unfortunately doesn’t make me coffee). I use it to clean up my messy meeting notes and summarize large chunks of text. I wish it were reliable enough to pull relevant and recent stats and sources from studies, as that would save me some Googling. Apparently, Perplexity does a better job of that, but I would probably end up verifying all the information myself anyway.”
“Working on my own, ChatGPT can also be a great brainstorming partner. I use it to help identify any gaps or potential improvements in my work, and to come up with synonyms, puns, and phrases that inspire me to come up with something unique.”
So, no, we aren’t totally against using AI. What we are against is outsourcing the role of content production to AI and reducing or eliminating the role of real human writers. There are other agencies out there making bank on generative AI content creation—we’re not one of them.
You may also be interested in: Our 14 best apps for small business owners
AI is average. We’re not.
AI can only remix content that has already been published elsewhere. That’s a meh from me. As a writer, as a human being, I want to read content that incorporates the subtleties of the unscrapable. Content that makes me learn something, and, more importantly, makes me feel something.
Get in touch if you want to stand above the rest with conscious content created with conscience. (Yes, it’s a mouthful. I’m human, after all.)
You may also be interested in: