The most dangerous word in marketing is “scale.”
“We’ll scale content creation to maximize awareness and optimize lead gen,” the CMO pleads to keep their CEO’s rabid second-guessing at bay. For two decades, producing high quantities of content was a simple way to create the illusion that you’re doing a good job. Create enough well-structured content on a topic, and eventually you’ll see a spike in search traffic, even if the content is meh. Pump out mediocre webinars and downloadables and enough dumb schmucks will fill out the form that you’ll hit your superficial lead goal and survive another quarter.
This is why generative AI makes CMOs' eyes glisten more brightly than an open bar at Cannes. It looks like an easy button for content creation. In 2025, we’re creating more content than ever, and 74% of it appears to be AI-generated. The only problem? This strategy doesn’t work anymore. High quantities of mediocre content are relatively worthless in the AI Age.
As the cost of creating mediocre content approaches zero, so does the value of that content. In 2025, mediocre content does little for your brand, particularly since bland, informational blog posts that answer questions like “What is content marketing?” simply don’t drive traffic anymore thanks to Google’s AI Overviews.
The B-minus content that was good enough to keep you from getting fired five years no longer flies. That’s why, in my talks and workshops to marketing teams about AI, I always start with a simple message: Beware the Vortex of Mid.
The AI Vortex of Mid
Large Language Models (LLMs) like ChatGPT, Claude, and Gemini are mid by design. (If you don’t speak Gen Z, mid means mediocre.)
LLMs learn to generate text from the most common patterns in their training data, which is comprised of everything our Silicon Valley overlords could scrape off the internet, copyright be damned. Most of that training data is poorly written — technical, jargony, and flat.
The LLMs are then fine-tuned through Reinforcement Learning from Human Feedback (RLHF for those in the know), in which humans rate the AI output, often in “digital sweatshops” in Africa and Southeast Asia. These human trainers are instructed to sway the AI towards safe, inoffensive outputs and rate jargony technical answers higher.
This is why AI writing hasn’t made the same leap as AI coding. With coding, AI can test its code until it produces something that works when deployed. With writing, AI systems are sabotaged by a training process that makes it sound like an insufferable junior McKinsey analyst from Connecticut named Brett. Sure, AI is better than the average human writer, but it’s far worse than the best human ones. AI is inherently an averaging out of all the shitty content on the internet, and as a result, it creates a vortex that sucks you back to the middle when you rely on it too heavily.
I call it the AI Vortex of Mid.
If you laid out all the content on the Internet on a 2x2 graph spanning Low-High Originality and Low-High Quality, most AI-generated or heavily AI-assisted content would fall in the middle. Not bad, necessarily, but not good enough to make a significant impact.
That’s due to the Vortex of Mid, AI’s strong gravitational pull towards mediocrity. Outsource your outline to AI, and you’ll be stuck with generic ideas. Outsource your prose, and it’ll lack soul. With the right prompting and fine-tuning, you might get the veneer or originality (AI is skilled at connecting ideas) and break into the upper-middle class of all content. But it won’t be enough to break through.
The honest discussion that marketing, content, and media leaders need to be having is about when Mid is good enough and when it’s not.
I’d argue that for most formulaic content (landing pages, nurture emails, that press release announcing your fifth pasty, middle-aged CRO in the last three years) mid is often fine. That press release isn’t ending up in the MoMA, and if using AI means your exhausted two-person content team gets to reinvest that time telling exceptional stories, the trade-off is probably worth it. Not every brand gets to be Oatly, tripping on irony and acid in every line of copy.
But you need to take that trade-off seriously and tell the kind of bold stories that AI cannot — ones backed by original research, reporting, and insight. If all you produce is Mid content, you’re fucked. And if you instinctively turn to AI every time you sit down to create content, you’ll see your critical thinking and communication skills atrophy. The Vortex of Mid will trap you, and you’ll never be able to escape.
If you liked this story, you’ll also like:
Chart of the Week
AI Search is truly the wild west. Case-in-point: ahrefs (whose content team is on a heater right now) found that 86% of top-mentioned sources are not shared amongst Perplexity, ChatGPT, and Google’s AI Overviews. As they write:
Google AI Overviews lean towards the same top sites we’re used to seeing in search results. There’s a heavy emphasis on authoritative websites for health, finance, encyclopedic knowledge, social media, as well as a bias for Google-owned properties. Notably missing are media sites, non-Google entertainment sites, and ecommerce sites.
ChatGPT leans more on publishers and non-Google media, especially for news, entertainment, and sports. They have licensing deals and partnerships with many of the cited websites. ChatGPT doesn’t have much around health or medicine.
Perplexity pulls from a broader international corpus. There are many local and regional brands that weren’t in the top 50 for the other AI assistants. Perplexity doesn’t have social websites or traditional Western media.
Recommended
The Three-Body Problem (Cixin Liu): I’m writing this from a balcony overlooking the sea in Sitges, Spain while on vacation with my family. As usual, I brought far more books than I could ever read while chasing my toddler around the beach, but this is the one novel I’ve been able to get through. It’s delicious and even better than the acclaimed TV series.
DraftGPT: The Brave New World of AI Hits the NBA (Jordan Teicher / The Ringer): This is an excellent piece by Jordan — a good friend and an even better reporter, who managed to get NBA teams to open up about their secret AI strategies.
Employees Are Buried in AI-Generated Resumes (Sarah Kessler / NY Times): Everyone who said AI will transform work was right! It’s now impossible to apply for a job.
I’m a Fractional CMO and the best-selling author of The Storytelling Edge. Subscribe for free to get new storytelling and audience-building strategies in your inbox each week.
"AI makes CMOs' eyes glisten more brightly than an open bar at Cannes" made me spit take
Excellent post as usual!
Thanks, this is timely. It's been less than a year that I've been tapping Chat GPT as a research and writing aid, and it just cannot write anything interesting and I'm feeling that pull to the Mid Vortex for sure. It synthesizes and generates great brainstorms. Very useful. But that dreaded Mid has crept into a lot of my writing for our new website, and I realize I've been sucked in and mesmerized by boring, safe writing with the same damn cadence over and over. I'm in the biz of video storytelling and my gut tells me that real human stories will have incredible currency as we go forward. But I can't add to the slop in my own written content. It's a sloppery slope.