Recently, I received quite an aggressive reaction to one of my comments. In that comment, I pointed out that employees of serious organizations have stopped paying proper attention to editing their content once it has been processed through LLMs.
The essence of the criticism I got was: If you talk about "not using AI," then why did you use an AI-generated image yourself? Doesn’t that contradict your point?
But this is exactly where my argument lies: there is a difference in how AI is used and in how careful the approach is.
This reminded me of the early days of web development, when IDEs started to appear. I remember colleagues who passionately argued that “true HTML coding” meant writing everything in Notepad and that using an IDE was “wrong” or even “harmful.”
Looking at today’s situation, I believe we are facing something very similar. And when it comes to LLMs (used for drafting articles, posts, and publications), the key to proper use is not simply generating content but asking yourself:
Does this output really meet the task I set?
Does it fit the context and preserve the original meaning?
That moment of revision, the act of carefully checking, editing, and polishing is where the author decides whether the result is worth publishing.
If the edited text or image aligns with your context, supports your message, and does not distort your original idea, then using AI is no different than using spellcheck in Word or autocorrect on your phone.
However, when people publish AI-generated material without even reading it through once, without editing or questioning it, that’s when doubts arise.
Is this truly your content that you shaped, or is it just meaningless text generated by a prompt?
We must remember: LLMs don’t create with intent or understanding. They generate statistically probable text. That’s why revision, human oversight, and critical thinking are what make the difference between quality vs. poor content, useful vs. useless results.
So when someone says, “If you advocate for responsible AI use, you shouldn't use it at all” — that logic is incorrect. (this kind of dash "—" is used by combination of “opt” + “shift” + ”-” on macOS :) )
Denying technology altogether leads nowhere.
The real success lies in understanding it, applying it responsibly, and striking the right balance. That is what makes any tool truly valuable.
In an era of tremendous content generation, human oversight is more important than ever. It seems we're close to the point where we'll begin creating content platforms that will aggregate "primary sources" strictly reviewed content.