Someone spends a week finding something out. You ask a question. A machine reads their work and gives you the answer in three seconds.
You have what you need. You close the tab.
That is a good experience. It is also where the problem starts.
For a large part of the web, discovery came with a visit. The page got a chance to show an ad, earn a subscriber, or introduce you to someone worth reading again. That arrangement also rewarded pages burying one useful sentence beneath eight paragraphs written for Google. Nobody owes those paragraphs their afternoon.
But an answer engine can remove the annoying part and the payment mechanism in the same interaction.
Pew Research studied Google searches from 900 US adults in March 2025. Traditional results got clicked in 8% of visits with an AI summary, versus 15% without one. Different queries trigger summaries, so that comparison alone cannot establish causation. Links inside the summaries got clicked in just 1% of visits with a summary.
People still click. Google said in August 2025 that overall organic clicks were roughly steady year over year. But an aggregate can stay steady while particular publishers lose their audience.
The question is what happens when being useful stops being enough to earn the visit.
Imagine a database company whose documentation helps an assistant fix a customer's error. The company may get exactly what it wanted: someone keeps using its product.
Now imagine an independent engineer who spent a weekend finding that fix. Their reward might have been a conversation, a job opportunity, or a reader who comes back. The answer can travel while those possibilities stay behind.
For a publication paying someone to investigate the problem, there is a salary to cover too.
A citation gives readers a way to check the source. It cannot guarantee they have a reason to visit it.
Wikimedia offers one alternative: charge commercial organisations for reliable, high-volume access while keeping the knowledge free. That helps fund the infrastructure. A small writer has less obvious bargaining power when another page offers a similar answer.
My worry is about the useful work that has no product behind it and no established paying audience. A local meeting somebody sat through. An obscure product somebody tested. An error that took two days to understand and takes one sentence to explain.
Telling those people to build a newsletter leaves a step unexplained. How does a stranger discover them? A newsletter gives an existing reader a way back. The first visit still has to come from somewhere.
I suspect the change will show up in what people bother publishing. Some observations may stay in private groups, where at least someone responds. Others may never get written down.
A good answer saves the reader time. Who is going to spend the week finding the next one?