Section I -- Situation Report
The Bureau of Information Saturation is pleased to confirm that global data generation reached 221 zettabytes in 2026 -- approximately 402 million terabytes per day -- representing a 22 percent increase over the previous year. All channels are active. All feeds are full. Every screen is lit.
Comprehension remains at prior-year levels. Which is to say: it remains absent.
Fifty-two percent of Americans now report feeling worn out by the amount of news there is. Only 36 percent follow the news closely, down from 51 percent in 2016 -- a 15-point decline in under a decade. Six in ten have reduced their overall news intake. Two-thirds have abandoned a specific source entirely. Only one in ten say they follow the news because they enjoy it.
Meanwhile, 74.2 percent of newly published web pages now contain AI-generated content. Nieman Lab projects that in 2026, AI-written output will outpace human production across the mainstream channels where people search, scroll, and learn. Europol experts have estimated that as much as 90 percent of online content could be synthetically generated by year's end.
The Bureau observes that the information economy has achieved an extraordinary condition: perfect saturation and zero comprehension, running simultaneously, in the same feed, on the same screen. This is not a paradox. This is the operating procedure.
Section II -- The Mechanism
The system works as follows. Platforms prioritise engagement. AI-generated content is cheaper, faster, and infinitely scalable. The content floods search results, social feeds, and recommendation engines. The volume increases. The signal-to-noise ratio declines. The user, confronted with more information than any previous generation and less confidence that any of it is reliable, makes the rational decision: they stop looking.
The Reuters Institute's Digital News Report confirms the trajectory. Across all markets surveyed, 40 percent of people now sometimes or often avoid the news -- the joint highest figure ever recorded. The reasons are consistent and unsurprising: negative mood impact (39 percent), feeling overwhelmed (31 percent), too much conflict coverage (30 percent), and a sense of powerlessness to act (20 percent).
The information is not scarce. The audience is.
Researchers have documented the cognitive mechanism. High information volume degrades decision quality, forcing the brain to fall back on simplified heuristics and cognitive biases. Eighty percent of workers now report experiencing information overload, up from 60 percent in 2020. Economists estimate the annual cost to the global economy at approximately one trillion dollars. The species has invested trillions in producing information. It costs a trillion a year because it cannot process the result.
BUREAU NOTE: The Bureau wishes to clarify that the information economy does not suffer from a shortage of information. It suffers from a surplus that has achieved the same practical effect. This is not ironic. Irony requires surprise. The Bureau has been expecting this since the invention of the printing press.
Section III -- Model Collapse (Or: The Ouroboros Has a Content Strategy)
In 2024, researchers published a paper in Nature documenting what they called model collapse: the phenomenon in which AI systems trained on AI-generated data suffer irreversible quality degradation. The tails of the original data distribution -- the rare, the specific, the genuinely informative -- disappear. The output converges toward an increasingly bland, increasingly confident, increasingly wrong centre.
The threshold is lower than anyone expected. Even a contamination rate of one in a thousand -- one synthetic data point per thousand real ones -- can trigger the collapse.
The practical consequences are already visible. AI-written pages in Google's top 20 search results nearly doubled between May 2024 and July 2025, climbing from 11 percent to nearly 20 percent. Zero-click searches -- queries in which the user never leaves the search engine -- rose from 56 percent to 69 percent. When Google's AI-generated summaries appeared, users clicked on actual results 8 percent of the time, compared to 15 percent without them. The search engine -- the tool built to help people find information -- is now serving machine-generated summaries of machine-generated content about machine-generated content.
The information economy is eating its own tail. The machines produce content. The machines train on the content. The machines produce worse content. The machines train on the worse content. The Bureau notes that this is not a metaphor. Nature published the paper. The feedback loop is documented. The ouroboros has a quarterly earnings call.
Section IV -- The Great Withdrawal
The most telling statistic is not the volume of content being produced. It is the volume of attention being withdrawn.
In 2016, 51 percent of American adults followed the news all or most of the time. By 2025, that number was 36 percent. The decline did not correspond to a reduction in the availability of news. It corresponded to an explosion. The species that built the largest information distribution system in history is now leaving the building.
The Pew Research Center's 2026 report on Americans' relationship with news captures the condition precisely. Nearly half of U.S. adults say most of the news they encounter is not relevant to their life. Among young adults, the disconnection is sharper still: Reuters found that 42 percent of those aged 18 to 24 sometimes or often avoid news. In the United Kingdom, under-35s are four times more likely to avoid news because they do not understand it than those over 35.
The audience is not failing the information economy. The information economy is performing within its stated parameters. It was built to produce content, not comprehension. It was built to capture attention, not to reward it. The Bureau notes these are different metrics and the correct one was always chosen.
BUREAU NOTE: The Bureau has consulted its records and can confirm that the information economy has never, at any point in its operational history, listed "understanding" as a deliverable. The product has always been volume. The customer has always been the advertiser. The reader was always the raw material. Any confusion on this point is itself a product of the system.
Section V -- The Machine Trains on Itself
The model collapse paper is the correct document for this moment.
The machines produce content. The machines train on the content. The machines produce worse content. The machines train on the worse content. The tails of the original data distribution — the rare, the specific, the genuinely informative — disappear from the training set. What remains is the confident centre: increasingly wrong, decreasingly varied, indistinguishable from the content it was trained to approximate.
One synthetic data point per thousand real ones is sufficient to trigger the collapse.
The Bureau observes that this is not a technical problem awaiting a technical solution. It is the information economy's operating biography written in advance. The system was built to maximise volume. Volume was maximised. The volume trained the machines. The machines maximised more volume. The comprehension was never in the specification.
BUREAU NOTE: The Bureau notes that Terence McKenna identified the absence of consciousness — not information — as the operative crisis, decades before the data confirmed him. The Bureau is not equipped to act on this observation. The Bureau can only file it, which it is doing now, in a document that will shortly contribute to the training set.
Section VI -- Bureau Assessment
All available channels are saturated. The machines are producing content faster than the population can ignore it. The quality of both the content and the audience's capacity to evaluate it are declining in tandem. Model collapse is not a bug. It is the system's biography written in advance. News avoidance is not a failure of civic duty. It is the market correcting for a product that no one requested and everyone received.
The Bureau will continue to monitor the situation. All feeds remain active. All channels are operational. Your capacity to absorb this report is, as always, your own concern.
BUREAU NOTE: You have now consumed approximately 1,500 words on the subject of information without understanding. The Bureau trusts you found them informative. Whether you found them clarifying is a question the Bureau is not equipped to answer. That has never been our department.
Bureau of Information Saturation — monitoring the gap between what is available and what is understood since the first library was built and the first book went unread. Your saturation is appreciated and on file.