The Content Creation Paradox: Why LLMs Need Us to Keep Writing

We’ve reached a point in history where LLMs are consuming everything on the web. These new products-whether it’s Google, Bing, or the various AI assistants-consume content that doesn’t really belong to them. They act as middlemen, delivering results without you ever accessing the original source. They eliminate the need to visit the actual page, to read the original author’s work, to see other people’s contributions on platforms like Stack Overflow.

At some point, we have to understand that new content has to come from somewhere.

This is a big mistake, in my eyes. The incentive to actually stay engaged on the web, to be geeky, to help out other people-that incentive is being demolished. And I think the people who decide to stay on the web and actually run their own websites, their own projects, their own platforms-they will win the race long-term.

The Platform Trap

Yes, crawlers love big, structured platforms. They understand Reddit, X, Substack, Medium. But these are not owned platforms. They control what you can post, whether you earn anything, whether anyone sees you. They control pretty much everything. Tomorrow, they can take away your account, block you, disable your payments, shut down entirely, or introduce yet another subscription fee. They have access to your analytics, your emails, your subscribers’ emails.

The main thing is just for us to understand: LLMs have to retrieve fresh data from somewhere.

What happens when people are no longer encouraged to share their experiences in written form? When everyone shifts to podcasts or videos only? Yes, writing is time-consuming. Yes, it’s difficult. Yes, it’s really difficult to do it right. But the AI we were promised-it needs fresh data.

The Token Collapse

At this point, they’ve already scraped the entire useful internet. Given the volume of content being generated daily, the next iterations of models will be predominantly trained on AI-generated content. The most probable tokens, squared. The language will become increasingly standardized, predictable, homogenized.

There’s no way forward but to enforce-to encourage-the creation of new, original content. Content that quite often doesn’t pay off, at least not directly.

I’m not really discussing Substack here, but I presume the situation is similar to OnlyFans: a really small percentage of people actually make money. And if we consider money the only incentive, we’re missing the point-especially for engineers. But we still have to encourage people to go the extra mile, to ignore the LLM results, to actually access websites and see the original.

The Value of Original Sources

The original might be more alive. It can have additional information posted by the author, values and context that get lost in summarization. It can be wrong too-but in that case, the person has their reputation on the line. They wouldn’t waste your time posting something stupid or useless just to destroy their own reputation.

With LLMs, we’ve forgotten what reputation actually means.

Why I’m Starting This Blog

This first post exists because I decided to have a static blog to post my articles, to host the components of my tutorials. And yes, I’ll admit it: this note has been transcribed from audio because I was too lazy to write it manually. Plus, it’s a markdown file-I need to understand how Hugo will render these files anyway.

But here’s the thing: this is still me, not an LLM generating tokens on my behalf.

We’re losing a lot by destroying the incentive for people to have their own media on the web. We’ve gotten stuck on corporate platforms that decide what happens to you tomorrow. At some point, we decided the convenience of unified platforms outweighs the risks-the risks of manipulation, exploitation, corruption, and generally bad products damaging the work of your life.

The Experiment

I want this static website to be indexed. I want to run an experiment to see how well LLMs index markdown files, whether my online profile will appear in answer engines, whether a static website can improve exposure by having fresh content that sits outside popular platforms.

Content that exists in the penalty box might actually have better value statistically-because it’s rare, because it’s an outlier. I’m not sure how answer engines judge the quality of content that’s not on a unified platform, that exists outside the standard structures. It’s probably difficult to understand what such a website is about, whether there’s a coherent content line.

In any case, it’s an interesting experiment.

I’ll start this blog and see how it goes. I won’t be posting frequently, but I can use it for my tutorials-n8n, Make.com, Zapier, and many other things. I’ll try to use this blog as much as possible and see how it performs in terms of Answer Engine Optimization.

AEO instead of SEO. That’s what they call it now.

Wish me luck