Unless you spend your free time studying economics, you’ve probably not run into the term “supply shock” all that often, if ever. So for you non-nerds, let’s start with a definition:
A supply shock is an event that suddenly increases or decreases the supply of a commodity or service, or of commodities and services in general.
You may be familiar with the concept — specifically a negative supply shock — when something dries up, causing prices to go wacky. Like the baby formula shortage in 2022 or masks during the pandemic or even the 1973 oil crisis (that last one being a very specific reference for the three boomers reading this article).
Which means I need one for the Gen Z crowd, I guess, so let’s go with the lack of semiconductors in the early 2020s making cars prohibitively expensive. Also PlayStation 5s.
Anyway, a positive supply shock, as you might imagine, is the opposite of that: when a whole lot of something floods the market.
Sound familiar? I hope so, because we need to talk about it.
Let’s Be Clear About What AI Content Is
First, we need to set a baseline for what is considered AI content, because right now, everyone is trying to make a black and white distinction inside a giant gray area. People are using vague descriptors like “tool” and “augment,” and trying to separate writers into either the Copy/Paste Clan or the House of Purists.
And that isn’t helpful. So let’s put it this way: If LLMs are saving you a significant amount of time, that’s AI content.
“That seems unfair,” you may be thinking. “The whole point of AI is to help us work faster and more efficiently!”
I know, but writing isn’t like software, and you can’t scale it the same way. Knowledge compression in people doesn’t function like data compression in computers. Regular folks can’t extract and process all the information that went into an LLM’s output, because it’s not a RAR file and human brains aren’t running a copy of 7-Zip.
Research is a critical part of the writing process, and you’re skipping it. It’s being outsourced to AI. That’s like reading the CliffsNotes instead of the book. You’re not doing the work.
So yes, if you’re using an LLM to come up with your ideas and your draft, even if you’re then editing and “humanizing” it, that’s still AI content. And the words generated by these language models are necessarily the most average thing ever, because that’s how the technology works. Let’s not forget that the output is, after all, “deepfakes for text.”
Meaning, you’re not writing — you’re just churning out “thin content.”
Which isn’t new, by the way. For you see …
We’ve Been Here Before
Let me take you back to the mid-2000s, when the internet was in its adolescence. Those awkward teenage years everybody hates talking about.
Google was establishing its dominance as THE search engine, and so began the grifty business of algorithm chasing. Which, if you don’t know, wasn’t producing well-structured, useful content, but instead figuring out what Google was rewarding, and just injecting the hell out of that into your website, artificially inflating your ranking in search results. There was even a name created for it: Adversarial Information Retrieval.
But in those days, instead of chatbots, we had content farms. An entire industry that popped up in order to produce mass quantities of crappy articles when word got around that websites ranked higher if they blogged regularly. And of course, since there was no AI, businesses just did the normal corporate America thing and outsourced it to poorer countries on the cheap.
Then, in 2011, Google introduced everyone to Panda.
If you weren’t around for the Panda update, you missed what I can only describe as Black Thursday for search rankings. Google was sick of all the gamification, and decided it was going to start checking the content it was rewarding to see if any of it was worth a damn. And if it wasn’t, they’d start doling out punishments.
Maybe “punishments” is underselling it a bit. It’s probably more accurate to say that Google completely decimated a shitload of websites all at once — primarily the most egregious violators, but the response was so severe that there was, in fact, some collateral damage.
But why would Google take such a heavy-handed approach, and not just with Panda, but with subsequent updates like Penguin, Hummingbird, Medic, and a whole list of others that followed?
Well, they had to. Because …
Search Must Be Able to Separate Signal from Noise
Before we talk about search, I want to dispel a myth. Specifically, this one:
AI search has taken over.
Look, I’m willing to listen if you have any non-survey evidence that shows 40% or 60% or whatever other crazy percentage of people are using AI for search. But right now, I’m seeing data like less than 1% of referral traffic, as of the time of this writing. To be fair, there’s been a noticeable dip in traditional search and an increase in direct traffic, which I’m sure is due to more people asking an LLM and then typing a web address directly into their browser. Absolutely.
But it’s not half the internet, and the reason is that people are still pretty wary about artificial intelligence. That’s why DuckDuckGo’s app installs went up 30% after Google crammed AI into their search bar.
Regardless, that’s not even the conversation to be having, because it doesn’t matter if it’s regular old search or a chatbot — the underlying information that either one is retrieving has to be real and accurate. It must be trustworthy, otherwise people will stop using it.
Google has been building an entire trust ecosystem for over two decades, and that’s the reason they have dominated the search space. The algorithm works wonderfully, Google Business Profile lets people know a place is real, and they look well beyond the website for establishing customer-friendly trust signals.
But in order to build and maintain this, they need to be able to separate signal from noise.
The internet is a big place, and the influx of AI-produced content has made it way bigger. There’s more information than ever for search (of any kind) to comb through in order to give users the results they want. And the system doesn’t work if someone can game it by getting an LLM to write a thousand blog posts on a topic. Anybody can do that now, which is why Google has made it a point to mention that they’re not a fan of automated content.
Simply having more words on your website isn’t the same signal it used to be, to traditional search engines, AI search, or humans. The bars for quality and value have been raised. And in order to understand why, we need only look at …
The Law of Supply and Demand
Don’t worry. We’re not covering Nobel laureate economics here.
For the purpose of this discussion, we just need to know the basics, like how when the supply of something outpaces its demand, the value goes down. We all understand and agree with that, right? Hopefully? Listen, if you’re not up to speed, go ask a kid with a Pokémon binder why nobody’s buying or selling Energy cards.
The point is, we all can create passable copy today using AI, for essentially zero dollars. But if that’s the case, how much is the output actually worth?
We don’t need terms like “supply curve” or “marginal cost” to ask some very basic questions here. Like, “if everybody can create content with almost no effort, how is Google going to judge it?” Or maybe, if you’re a marketer, “if a business owner can get Claude to write blog posts, why are they going to pay me to do it?” Or this one: “If anybody can whip up a newsletter in two minutes, why would my customers care enough to read mine?”
Even if it were true (it’s not) that “nobody cares how it was created,” and that the result is the only thing that matters, LLMs still cannot write the way people do. They might be able to write better than you, but that’s not the same thing, and you may want to be careful about how hard you defend that position on LinkedIn. Because as we’ve established, the only thing AI does is spew out probabilistic results. That’s its whole job.
Since anybody can wield a chatbot to produce content at scale (which, fair warning, can end in catastrophe), it means that differentiation is the name of the game, and that’s something LLMs just can’t do. By design.
Sure, AI might be able to save you time, but what is that content doing for you? What, exactly, are the LLMs creating? Because right now, the answer to that last question seems to be: the exact same thing as everyone else.
And that’s not very sellable.
