The Infinite Buffet with a Locked Kitchen
I spent yesterday morning chasing a specific thread about peat bog restoration in Scotland. Normally, this involves a delightful rabbit hole of specialized blogs, PDF reports from field researchers, and local news outlets that have covered the same ten miles of marshland for decades. Instead, Google handed me a tidy, three-paragraph summary that felt like a lukewarm cup of instant coffee. It was efficient, sure, but it felt hollow. It told me the 'what' without any of the 'who' or 'how.'
This is the information grazing paradox. We are currently witnessing a 40% drop in referral traffic to the very sites that provide the raw data for these AI summaries. It’s like a library that summarizes every book for you at the front desk so you never have to walk through the stacks. It feels like progress until you realize the library hasn't bought a new book in three years because nobody is visiting the shelves.
I wonder if we’ve forgotten that information isn't a natural resource like sunlight; it’s a manufactured good. It requires boots on the ground, cameras in the mud, and journalists who are willing to sit through four-hour town hall meetings about water runoff. When the AI scrapes that labor and presents it as a free-floating fact, the economic tether between the creator and the reader snaps.
The Mathematical End of Niche Truth
Let’s look at the numbers because they are genuinely startling. When a niche environmental site loses 40% of its traffic, it doesn't just lose 'clicks.' It loses the ad revenue, the newsletter signups, and the donor leads that keep the lights on. For a site like Grist or a specialized data-heavy climate tracker, that 40% represents the margin between staying solvent and shuttering the operation.

Photo by Engin Akyurt on Pexels
If these sites go dark, what does the AI summarize next year? It’s a snake eating its own tail. We are training sophisticated models on a corpus of knowledge that we are simultaneously defunding. I find myself wondering if we are accidentally building the most advanced echo chamber in history—a system that can synthesize every existing thought perfectly but has no way to ingest a new one because the observers have all gone out of business.
There is a specific kind of 'ecological truth' that only comes from longevity. You can't summarize thirty years of soil data if the person who has been measuring the soil for thirty years can't afford their hosting fees. We are essentially strip-mining the intellectual landscape, taking the topsoil of facts and leaving a desert of unverified AI hallucinations behind.
The Cost of the Convenience Tax
We always pay for convenience; we just rarely pay in cash. In this case, the 'convenience tax' is the loss of nuance. When I read an AI summary about a new carbon capture technology, I miss the skeptical quote from the local engineer or the photograph showing the actual scale of the machinery. I get the 'truth' but I lose the 'context.'
- The Ghosting Effect: Specialized sites are seeing 'zero-click' searches skyrocket, where the user gets the answer and leaves.
- The Training Lag: AI models are trained on historical data, meaning they are inherently backwards-looking.
- The Funding Gap: Philanthropy cannot replace the $30 billion digital advertising ecosystem that currently supports niche journalism.
I'm curious if we will see a return to the 'walled garden' era, where the best environmental reporting is hidden behind aggressive paywalls or moved to private Discord servers. If you can’t stop the AI from scraping your work, you stop making your work public. That would be a tragedy for the climate movement, which relies on the free flow of data to hold institutions accountable.
What This Actually Means
We are currently participating in a massive, unplanned experiment in information ecology. By prioritizing the 'answer' over the 'source,' we are effectively telling the people who find the answers that their labor is worthless. If the current trend of a 40% traffic decline holds, the 'information grazing' habit will eventually lead to starvation. The AI will be left summarizing itself, a feedback loop of increasingly degraded data that lacks any anchor in the physical world.
I don’t think the solution is to ban AI summaries—the genie is out of the bottle and it’s very good at summarizing things. But we need to find a way to credit the 'intellectual supply chain.' If a summary relies on a specific investigation by an independent journalist, that journalist should be compensated, not just cited in a tiny footnote that no one clicks.
Ultimately, I’m left wondering what the internet looks like in 2030 if the only people left writing are the ones who don't need to be paid. We might end up with a digital landscape filled with corporate PR and AI-generated filler, while the actual truth about our changing planet becomes a luxury good, buried in private databases far away from the reach of a search bar.
Quick Answers
Is AI actually stealing content?
It’s more like a forced synthesis; it takes the value of the information without providing the traditional 'payment' of a website visit.
Why can't these sites just find new ways to make money?
Niche journalism has already spent a decade pivoting to every possible model; losing 40% of their top-of-funnel traffic is a blow most cannot survive.
Does this affect all news equally?
No, it hits niche, data-heavy sites hardest because they provide the specific 'factual' answers that AI is best at summarizing.



