AI Writing Has Likely Plateaued, and That’s Good for Brands

Hot take: there’s a good chance that AI’s writing abilities have already peaked, and they’re… just okay. 

Maybe not a surprising take, from a writer, but hear me out. My skin in the game aside, scientists, critics, and even Bill Gates have been questioning whether AI can get much better since at least 2025. AI lovers and haters alike claim that artificial intelligence’s abilities are exponentially growing. But, at least when it comes to LLMs, that may not be true. 

Why have AI’s writing skills stalled out? The technological and economic reasons behind the intelligence slowdown are numerous and complex — I could write a book. But these are just a few. 

Chat’s Inspiration Has Run Dry
It’s been said that ChatGPT has “read the entire internet,” though this isn’t exactly true. AI corporations are cagey about publishing real stats on how much their models have consumed, but what we do know is that major LLMs have read trillions of words in many mediums, from novels to online forums to news articles, to the chats that millions of users engage in each day.

In short, it’s pretty much consumed and learned from as much content as it possibly can.

Even if LLMs continue to read and learn from the internet, there’s a major problem: AI content has now surpassed that written by people. As LLMs continue to scrape the internet, they are now increasingly consuming their own work, like a snake eating its tail. 

And on that note, bot traffic is also surpassing human traffic. For us human marketers, this is bad. It skews our understanding of what actually resonates with real people and what’s resonating with AI. 

For AI learning itself, it’s even worse. 

A Dead Internet Made for Robots, by Robots 
Enter the Dead Internet Theory, aka the cycle in which AI creates content. That content is then trafficked and “liked” by AI bots, which then causes AI content creators to cater even more to said bots, neglecting the preferences of actual humans. 

In certain dark corners of the internet (Facebook), the Dead Internet phenomenon has already come to pass, e.g., the ouroboros that created Shrimp Jesus in 2024. If I’m being honest, I miss this era of AI. If we must succumb to robots, let it be ones that worship a Boschian crustacean overlord. At least it’s a creative ending.  

But no. Unfortunately, it’s more likely the internet will just become an echo chamber of mid-ness. Corporations have reigned in their AI models since the days of Shrimp Jesus, making them safer for brands to use and investors to invest in. The key word being not safe, but safer — LLMs still don’t have the common-sense reasoning skills and nuance a human does to avoid, say, racial bias, or talk about tricky topics. But I digress.

Why AI’s Creative Plateau Is Good News for Brands
When it comes to creativity, LLMs’ unpredictability (read: originality) has been checked by AI companies in order to produce “safer” creative for brands. In turn, LLMs have become less creative and, in my experience, worse at writing truly original content than they were a few years ago. Because, as any good creative knows, safe means boring. 

For many who work in advertising and marketing, the plateau of AI’s writing skills might sound disappointing at first. But savvy brands will use this to their advantage. No, LLMs won’t be able to reliably produce solid, exciting, nuanced, or daring writing. People always will. 

Regardless, brands will continue to take shortcuts and use AI to write their ads and content anyway. Great. In a sea of sameness, it will be easier than ever for smart brands to stand out with content and copy that’s unique and human, that pushes boundaries and tests new ideas. 

Let’s Use AI for What It’s Actually Good At
This isn’t to say that AI can’t keep improving other areas of our work. There are skills it excels at, just not this one. 

As AI is integrated into more and more work processes, the best we can hope for is that it will be a valuable tool for handling tedium, so that we can focus on work that requires big thinking, dreaming, risk-taking — and writing.