As AI-generated search answers increasingly deliver information without requiring users to visit the websites that produced it, a larger question is emerging: who pays to create, maintain and preserve the public knowledge AI depends on?
WHAT’S HAPPENING
A seemingly simple problem with Google displaying an inaccurate sunset time has become part of a much larger debate over the changing structure of online information.
Search engines are increasingly moving from directing people toward websites to generating answers themselves. That can make finding information faster and easier, but it can also reduce the number of people who visit the original sources.
The concern extends well beyond Google.
Wikipedia, the Internet Archive, libraries, government databases and independent publishers all contribute to an enormous digital knowledge infrastructure that people — and increasingly AI systems — rely upon.
English Wikipedia pageviews, for example, have declined from roughly 7–8 billion per month in early 2024 to around 6–7 billion per month in 2026. The reason for that decline cannot automatically be attributed entirely to AI, but it is occurring as AI-generated answers become a larger part of how people retrieve information.
At the same time, governments and institutions are paying greater attention to digital sovereignty — the ability to maintain greater control over the technology, data and information infrastructure their societies depend upon.
WHY IT MATTERS
AI doesn’t create its knowledge infrastructure from nothing.
It depends heavily on information produced by researchers, journalists, governments, encyclopedias, archives, businesses and millions of websites.
That creates a potential structural problem.
If AI systems extract useful information from those sources while fewer people actually visit them, some of the organizations responsible for creating and preserving that information could receive less traffic, revenue, attention or financial support.
The result could become a feedback loop:
AI becomes better at answering questions because the open web exists — while simultaneously reducing some of the incentives that help keep that open web healthy.
This doesn’t mean AI search is inherently harmful. Faster answers and easier access to information can provide enormous benefits.
But it does mean the economics of public knowledge may need to evolve alongside the technology used to retrieve it.
WHO BENEFITS
Users gain faster access to information without navigating numerous websites.
AI and search companies can provide more complete answers directly within their platforms.
Governments and institutions investing in domestic digital infrastructure may gain greater control over critical information systems and reduce dependence on a small number of foreign technology providers.
Organizations that develop new licensing, attribution or compensation models could also benefit if the information economy begins placing greater value on original sources.
WHO LOSES
Publishers and independent websites could lose traffic when AI answers questions without requiring users to click through.
Public knowledge organizations could face additional financial pressure if declining traffic eventually translates into declining awareness, donations or institutional support.
Researchers and future generations could ultimately lose access to portions of the historical digital record if websites disappear faster than they can be preserved.
And AI companies themselves could eventually face a problem if the high-quality information ecosystem used to train and support their systems deteriorates.
WHAT HAPPENS NEXT
Expect the debate over AI search to expand beyond accuracy and hallucinations.
The next phase is likely to involve questions about attribution, compensation, archival access, public digital infrastructure and national control over information systems.
Europe is already showing growing interest in technological sovereignty. In June 2026, French search engine Qwant said it became the default search engine on European Parliament internal computers — another indication that control over search and digital infrastructure is becoming a strategic issue rather than simply a consumer preference.
The larger question may therefore become less about whether AI can answer our questions.
It may be:
Who maintains the knowledge underneath those answers — and what happens if the institutions doing that work can no longer afford to?
