I spend so much time relying on various AIs that, when I talk with other SME management consultants, I sometimes wonder if there’s anyone more shamelessly “all-in” on AI than me.
And yet, there’s something I’ve realized along the way. I’ve walked alongside AI’s growth for almost three years now, and after analyzing a wide range of data, there’s one thing I can say with confidence today:
Content written by AI doesn’t move the numbers.
Today, I’d like to dig into why this happens.
AI isn’t catching the “latent keywords” behind people’s and companies’ problems
Needless to say, AI feeds on information from the internet—websites, social media, and everything in between.
With so much information online, you might think, “If we just scrape it all together, can’t we solve any problem?”
People face all kinds of challenges every day. Some are simple; others are complex.
And the more complex a problem is, the more likely it is that multiple causes are intertwined within that one issue.
If a problem stems from causes A + B + C, AI can instantly surface A, B, and C, present them neatly, and even propose solutions.
That’s exactly the everyday convenience we all enjoy when we say, “AI is so useful!”
But what about cases like this, where the causes look more like:
“It might be A, but also B + yesterday it was C, today it’s D + you could argue E + maybe F and G too + H is probably a given? + in case XY it could happen, but not Z… or maybe Z after all.”
In reality, the stubborn problems that individuals or organizations can’t seem to solve are often built from this kind of messy, shifting mix.
“Then if AI can’t solve it, Google search won’t either, right?”
Exactly. Not every problem can be resolved on the internet.
However—while you’re investigating that long chain of “maybe A but also B, yesterday C, today D…,” you might stumble upon a piece written by a true expert and encounter a keyword “K.” And in that moment you feel the jolt:
“So it was K. K was the real cause!”
Human-written words can have that kind of power.
Supporting many clients with content creation, I’m often surprised by these keywords that only surface from someone who’s genuinely done the thinking and lived the field. Readers come away moved: “I can’t believe someone out there wrestled with the exact same issue as me!”
AI is wonderfully convenient—I’m impressed by it every day—but I’ve never felt that kind of emotional jolt delivered by AI.
The internet is a democracy — numbers rule
For better or worse, the internet has a strong pull toward the mean.
Click-through rates, watch time, bounce rates—algorithms optimized for metrics tend to elevate expressions that the majority can accept without friction, and AI-generated prose fits that mold perfectly.
What gets lost in the process are the words that precisely name the pain of the few.
“Latent keywords” like K rarely sit in the dead center of the average; they hide on the edges of the field—among exceptions, bug tickets, and the footnotes of complaints.
If you’ve studied statistics, you already know: an “average” doesn’t mean many people sit exactly at that value.
Solving the myriad problems that fall outside the average requires something else entirely.
I suspect that’s why AI-written blog posts often feel oddly skimmable—but not worth truly reading.
And yet, there’s no going back to a world without AI
By now, some of you may have noticed:
A portion of the paragraphs above was generated by AI—specifically, the lines starting with:
“For better or worse, the internet has a strong pull toward the mean.
Algorithms optimized for metrics… elevate expressions the majority can accept…”
I fed the preceding text to AI and asked it to continue. It understood the thrust of the piece and connected the dots quite well. It really is remarkably capable.
This is how I operate: I gently criticize AI at times—and keep using it anyway. I weave in my own perspective while leveraging AI, and I use AI while structuring around my own perspective.
To me, AI is a partner we co-evolve with.
There’s no going back to a world without it.