Thoughts on LLM Integration Features

I never intended to release this blog engine from the start.
I was just trying to build a clean blog for my own use when things grew bigger, and looking back, the reason I decided to publish it was quite naive and reckless.
I had thought way too simply, assuming, "If this engine does well and gains traction, maybe I can build in mechanisms to naturally suppress or filter out the junk posts flooding search results."
In short, that innocent rationale of creating a clean blog ecosystem free of spam posts was the starting point of this engine's development.

My blog engine is structured to support unlimited multilingual expansion.
Right now, when I write in Korean, it generates English and Japanese posts along with URL slugs via LLM translation.
But as I was adding a feature to automatically generate English slugs from Korean titles, a sudden reality check hit me. I started this project because I hated spam posts, yet I was using an LLM myself.
I realized that no matter how well I optimized the structure and SEO, if others took this engine and ran LLMs to pump out junk posts in every language across the globe, I would have no way to stop it.

If someone is determined to bypass the setup using external automation scripts to push posts, I can't completely block them anyway.
In fact, I intentionally made integration annoying and tedious, but I wonder if opening it up as a convenient posting feature in tune with the times is the right move.
The text LLMs write these days isn't just simple keyword stuffing like before; it looks plausible on the surface. But because of hallucinations, it ends up producing another, even more serious kind of junk—plausibly packaged fake information.
They say LLM API or MCP integrations are the trend, but making it convenient with my own hands feels like going in the exact opposite direction of what I originally set out to do.

Of course, framing it simply as 'LLM usage = generating junk posts' is an oversimplification.
After all, an LLM is just a tool used by humans, and people use it to achieve their intended goals.
The core difference lies in the user's 'intent.'
The issue is that an LLM itself cannot distinguish whether the intent is merely to farm traffic for economic gain or to smoothly refine accurate information and arguments one wants to convey.
Ultimately, it comes down to the user rather than the tool, and there's really nothing that can be done about it.
Besides, if someone is set on doing it, automating a post-factory mode is entirely possible even under the current setup.

To be honest, practical thoughts cross my mind too.
Interest in this engine right now is practically zero.
I sometimes think, 'If it's going to fade into obscurity anyway, should I just attach LLM automated posting features to draw some attention?'
But then I get hit with a reality check, wondering why I bothered making this in the first place.
On the other hand, I feel like dropping it altogether, thinking, 'No one else uses it anyway, so why am I overthinking this for something only I use?'

In the end, it feels like I'll eventually have to integrate it to keep up with the times, but I'm still debating how far I should open it up.
Blocking it completely feels like isolating myself, while opening it up entirely feels like wiping out the whole point of creating the engine, leaving me unable to reach a conclusion.
It's a dilemma with no clear answer anyway, but for the time being, I plan to sit with this ambiguity a little longer.