Free Words, Expensive Eyes

In 2026, the experience of staring helplessly at a blank screen has all but vanished. A single prompt conjures polished blog posts, reports, newsletters, and proposals in seconds. Everyone calls this efficiency.

Yet something is off. Writing has never been easier — but being read has become nearly impossible.

The word 'efficiency' is always used to mean reducing costs. That is true. AI has driven the cost of producing a single piece of writing to virtually zero. The problem is that this cost has not disappeared — it has simply shifted.

When production becomes free, the scarce resource shifts from the content itself to the attention it must compete for. Writing multiplies without limit, but a person's day is still 24 hours and their eyes are still two. When supply explodes, prices spike somewhere else.

So we need to reframe the question: whose costs has this technology reduced, and whose has it increased?

The winners are obvious: those who churn out content by volume, marketers angling for top search rankings, platforms that sell advertising on traffic. For them, infinite production is a blessing.

The cost has been offloaded onto the reader. The labor of sorting the real from the fake, the verified from the plausible fabrication, has been transferred wholesale to the audience. Work that editors and news desks once performed is now done by each of us, every day, for free.

Here it is worth examining the role of platforms. Recommendation algorithms select for writing that holds attention, not writing that is good. AI-generated content is optimized precisely for that metric. As a result, feeds converge toward the mean — everyone saying similar things in similar tones.

Some researchers call this model collapse. AI learns from AI output, and that output becomes the next round of training data — a self-reinforcing loop. The endpoint is not abundance but a swamp of averages. Information multiplies while signal diminishes.

This is what makes certain things remain costly to the end: curation, trust, human endorsement. Someone putting their name on the line and saying, 'I read this, it is true, and it is worth your time.' What machines cannot replicate is not the sentence — it is the accountability.

A strong counterargument presents itself: doesn't good writing rise to the surface eventually? The belief that quality wins.

Only half right. Once noise exceeds a certain threshold, quality gets buried before it is ever found. It is not good writing that wins — it is the person with a trusted channel to surface good writing. Those are two very different things.

In Busan, this dynamic cuts sharper. The voice of a city that is not the center was already faint. Now it is buried once more beneath a flood of mass-produced content out of Seoul. Algorithms are designed to amplify what is already loud.

Paradoxically, that is precisely where the opportunity for the periphery lies. When everyone moves toward infinite replication, the things that cannot be replicated — concrete, place-bound witness and the name of the person who vouches for it — become the true scarce goods. The louder the noise, the higher the value of a real voice.

The argument is not to halt the technology. AI-powered production cannot be undone, nor should it be. But who gets to write the rules for how this tool operates has not yet been decided.

Rules can be made: mandatory disclosure of AI-generated content origins; transparency about what recommendation algorithms are optimizing for; structures that place economic value on the labor of human verification. None of this is a rejection of technology — it is a design choice.

In the end, there is one question that matters: in an age when anyone can write without limit, who do we let decide what gets read? Whether we hand that power to traffic algorithms or leave it with people who put their names on the line. Efficiency is already in our hands. What remains to be determined is the rules.