Every day, millions of Americans open a browser, type out an opinion, answer a stranger’s question, or recommend a product on a public forum, and then close the tab and think nothing of it.
They are not selling anything.
They are not being paid.
And yet, the words they leave behind have quietly become one of the most valuable raw materials in the entire technology industry.
The dollar figure attached to that casual conversation is only now coming into full view, and it is far larger than almost anyone guessed.
The internet has always run on free labor, but something just changed
Since the earliest message boards, the social web has been built on a simple, unspoken bargain: people write things, platforms host them, and everyone benefits from the exchange of ideas.
No money changes hands between the person typing and the platform.
That arrangement worked smoothly for two decades.
Then a new kind of buyer arrived, one that did not just want to read the conversations, but needed to consume them at scale to train large language models capable of answering any question a human being might ask.
Suddenly, the supply of authentic, free-flowing human text looked very different from the inside of a boardroom.
The stakes of that shift are still accelerating, and the numbers have grown far beyond what most users imagined possible.
Why a forum post trains an AI better than almost anything else
Academic papers, news articles, and corporate press releases all have their place in an AI training set.
But they share a common flaw: they are polished.
Forum posts are not.
They are written the way people actually think and speak, with tangents, corrections, disagreements, slang, and the kind of hyper-specific local knowledge that never makes it into a Wikipedia article.
That texture is exactly what AI developers need to build models that feel natural rather than robotic.
The demand is not slowing down either.
Epoch AI, a research organization that tracks AI scaling constraints, projects that the stock of publicly available human-generated data will be fully utilized between 2026 and 2032.
The indexed web contains roughly 400 to 500 trillion tokens of unique text, yet frontier models consume data faster than the internet produces it.
A market that barely existed three years ago is already worth billions
For most of the internet’s life, training data was simply scraped, taken without payment from whoever had published it.
That era is ending fast.
The global AI training dataset market was valued at $3.59 billion in 2025 and is projected to grow to $4.44 billion in 2026, according to Fortune Business Insights.
The growing recognition that data quality, rather than model architecture, is the primary determinant of AI performance is repositioning training data from a cost center to a strategic competitive asset.
Publishers that once gave their content away for free are now sitting at a negotiating table, and the biggest platforms are starting to understand what they actually hold.
The dynamic is similar to the one playing out in AI inference, where, as our colleagues reported, open model partners are citing up to 20x cost advantages over closed systems, a sign of just how sharply the economics of AI are being recalculated across the board.
The $200 million secret hiding inside everyday forum threads
The platform sitting at the center of this story is one most Americans use without a second thought: Reddit.
In its IPO prospectus filed in early 2024, Reddit disclosed data licensing arrangements with an aggregate contract value of $203 million, with deal terms ranging from two to three years.
In February 2024, the same day it filed its IPO prospectus, Reddit announced a content-licensing deal with Google for sixty million dollars a year, giving Google access to real-time content from Reddit’s vast user-authored forums.
A few months later, Reddit struck a similar partnership with OpenAI that is estimated to be worth around seventy million a year.
Analytics platform Profound has found that Reddit ranks among the most cited domains across major AI answer engines, including Google AI Overviews, Perplexity, and ChatGPT.
Reddit CFO Drew Vollero has described what drives that demand. “What makes Reddit’s data particularly valuable is its scale, freshness and commercial intent,” he told Benzinga, adding that the company is home to one of the “largest collections of human conversation on the internet.”
Reddit is discussing with Google and OpenAI a future deal structure that could enable dynamic pricing, where the platform earns more as its data becomes more valuable to AI-generated answers.
Reddit CEO Steve Huffman said during a 2025 earnings call that “every variable has changed since we signed those first deals” and that the company’s corpus is “bigger” and “more essential,” putting it in a strong position to renegotiate.
The same shift is reshaping how AI companies think about every kind of digital content, including the 25-year image archive that platforms have accumulated without most users ever realizing it.
What this means for the people who typed those posts for free
The honest answer is that the question of who benefits is still being worked out in real time.
These talks are unfolding while the broader fight over AI training data heats up, with some publishers suing AI companies for using their work without permission.
The New York Times has taken OpenAI to court, while Reddit itself has sued Anthropic for allegedly scraping its forums without authorization.
New frameworks are emerging to address the imbalance.
Reddit joined Yahoo, Medium, People Inc., and other companies in backing Really Simple Licensing (RSL), which aims to standardize how AI developers license content and compensate publishers, modeled after music industry frameworks like ASCAP and BMI.
The deeper point is simpler: every answer typed into a public forum has always carried more value than it appeared to.
AI just finally put a price tag on it, and that price turned out to be in the hundreds of millions.
The next time you answer a stranger’s question online, you are participating in something far larger than a conversation.
