What are we doing with AI?
In early 2025 I learned to build custom GPTs and had a fantastic time playing with large language models (LLMs) as productivity tools. I built a Notion workflow to manage client projects and book editing deadlines. I earned an advanced prompting certificate from Vanderbilt and showed clients how to prompt their way into better quality outputs.
I created a daily accountability workflow that helped me launch two new brands from scratch and finally publish a book I’d been working on for 15 years.
It was like having another me to bounce ideas off of, and it behaved like a Gen-Z assistant with an adorable sense of humor. Generative AI was a little buddy, cheerleading me across daily finish lines.
It was a giddy, fun time to experiment with LLMs.
Then something happened. No, it was the convergence of three things:
ChatGPT 5.0 dropped, and the internet lost its mind. Legacy media missed the point on why it was such a massive shift (spoiler: it wasn’t because people’s companions didn’t work anymore).
Model usage hit the tipping point and suddenly all my clients and all my friends and all my colleagues were generating content all the time. (Token limits suddenly became a thing.)
Enterprise models were mainstreamed into corporate workflows, which slowed down the systems for all of us. And across the board, copy was suddenly flat, bland, and forgettable.
It was almost as if, overnight, every client and every agency had decided they could prompt their way into the future.
At first, I went through the five stages of grief over losing my career. Surely this was the end.
I spent so much time making jam and pickling vegetables that my parter suggested I start selling them at farmer’s markets. (Ha.)
I envisioned a future where I raised alpaca and spun yarn on a homestead where I could grow my own veg after the digital apocalypse.
I made peace with the fact I would be a writer for fun, because surely AI would outperform me soon.
Yeah. Then 2026 rolled around. Ha-ha.
Just as quickly as AI hit the mainstream, the edges started fraying and clients were boomeranging back after embarrassing content fails.
Memory limitations. Running out of tokens. The rising cost of subscriptions.
Not to mention all the AI “tells” and hallucinations — editing the output was too much for internal teams to manage. I got more work editing AI output than I had when I was doing copywriting from scratch.
Now, when I hear people talk about AI taking our jobs or replacing creatives, I think, “when we get infinite computing power, infinite energy production, and infinite time to fact-check AI output … then I’ll worry.”
Where are we going with AI?
Let’s go back to that GPT-5 model update. Yes, some people were mad because their AI spouses “broke” and dumped them. It’s not really funny, but I admit to rolling my eyes and laughing a little when that happened.
What really broke was something more important (to me, anyway). All the custom GPTs I had built to that point — the ones that could differentiate between brand voices and hold memory from one chat thread to another within a project — no longer worked.
On a practical level, I went from being able to generate on-brand blog post and video script drafts for 3-5 specific brand voices with different style guides to having to re-train a lobotomized model every time I needed content.
It can’t be understated how frustrating that was. Before the switch, I was still doing my own editing. I was still polishing and fact-checking my output. But prior to the change, the output was decent to good quality.
That change effectively made the model useless to me.
I know, I know — boohoo, I had to go back to manually writing from the blank page. I can do that. It wasn’t a huge deal. But overnight I went from being able to produce 8-10 pieces of unique content per week to maybe three or four.
It was a significant drop in quality.
And now? This is just my observation of my own workflow, but hallucinations seem more frequent now. Again, this is only my opinion and observation of my own experience. I keep reading hallucinations are down and quality is higher now — that could be true. Some users may be having great results.
I can only compare my own output today versus a year or two ago. For me, there was a quality drop.
Will that reverse? Maybe.
Do I still use generative AI? Every day. Just not in the same ways.
Here’s where the models seem to shine:
Image generation
Outlines (here’s what happened when I used an LLM to outline a novel, though)
Challenging assumptions (when you specifically prompt it to challenge you)
Providing audience insights (it can scan forums for trends)
General research (if you tell it to provide source links and actually check them)
First drafts (as long as you’re specific and double check claims)
No, really, where will we end up with AI?
Let’s just say I’m not renting that farmer’s market booth just yet. Still buying my alpaca yarn.
I used to believe we’re in the middle of the next industrial revolution. Maybe we are. I’m not so convinced anymore, though. I use the tools too much to have infinite confidence in them.
Language models are extremely useful. They shave time off of ideation and even research (please fact-check).
I’ll continue using generative AI for all kinds of things. I’m not a doomscroller when it comes to AI announcements or news.
But will it replace human editorial judgement? No. Not in my opinion. It doesn’t understand human context or nuance. It confidently invents information. People can sense when they’re experiencing generated media and they tune out.
AI backlash is real. Brands and writers who rely on LLMs to produce content still need to humanize it and preserve their individual voices. Why is user-generated content so popular? Why do average popular YouTubers pull in bigger audiences than dying legacy media shows?
Human authenticity.
AI will disrupt many things across many industries. However, what I’m seeing so far in content production is a whole lot of errors and people turning their backs on AI-generated content. Models would have to get a whole lot more accurate, less expensive, and more human before they could take over human-generated content.
My ten cents, anyway.