In November 2025, Scott Lapatine, founder of the independent music site Stereogum, said it paid freelance writers within a week of receiving an invoice. His appeal to readers also noted the nearly 200 new voices the site had introduced. He said it did not take unpaid work. Before a music story becomes something a search engine can find or an AI system can summarize, somebody has to do the work, and somebody has to make paying for it a priority.
The web is getting very good at moving information away from the places that produce it. That can help a reader. It can also break the links through which a source finds money, new writers and reasons to keep going. The danger in a zero-click web is that answers are still easy to find while the work needed to keep them useful becomes harder to sustain. For the web to stay useful, its sources need support. They also need people who have yet to write their first story or make their first edit.
Original information has to be renewed
A page can look finished for years. The work behind it often cannot. A product changes. A public official leaves office. A finding is challenged. A restaurant closes, a rule changes, an old account no longer explains how something works. The old page is still there, neatly laid out and easy to quote.
Keeping it useful requires someone to notice what has changed and decide what to do about it. Sometimes that means a correction. Sometimes it means starting again. The cost does not disappear because an answer service can put an answer together in seconds.
Reporting faces the same problem before there is a page at all. A review needs someone to try the thing being reviewed. An investigation needs evidence that somebody has found and checked. AI can help with that work, but a fluent account does not prove that someone saw or tested what it describes. A thousand summaries of one test still depend on that test.
Lapatine described how hard it had become to fund this work. He blamed AI-driven search changes for a 70% fall in Stereogum’s ad revenue that year, despite efforts to find other income. That was one publisher’s explanation, not a controlled estimate of AI’s effect. Earlier social-platform changes were part of his account too. It should not become a forecast for every website.
Even as the total number of pages grows, a source may struggle to pay people for the work others rely on.
A source can remain online while doing less. Answer services may keep retrieving its archives, and readers may still trust its familiar name. If a site can no longer afford to check or update its work, readers may see no broken link to warn them. The loss may appear much later, when a subject has changed and the source has not kept up.
A visit can bring more than ad revenue
Wikipedia makes the problem harder to dismiss as a complaint about advertising. It aims to share knowledge widely, including beyond its own website. When someone learns from its work elsewhere, that can serve its purpose.
In October 2025, Wikimedia’s Marshall Miller reported that recent human pageviews were about 8% lower than in the same months of 2024. Changes in how bots were counted made the figures harder to compare. AI answers and changing social-media habits were thought to play a part, though their effects had not been measured in isolation. Fewer readers could also mean fewer people who might volunteer or donate.
A reader might join in the work. Someone who spots a gap might fill it. Someone who returns to the site for help might choose to support it. Most visitors will do neither, but an open project still needs routes through which people can begin. When another service gives readers the answer without showing them the source or its community, they may be less likely to get involved.
This does not prove that AI has already reduced the number of Wikipedia editors, much less that the site is about to disappear. But sharing its knowledge more widely could leave fewer readers in contact with the people who keep it up to date.
A check alone cannot bring in someone whose skills are missing or teach a new contributor how to make a useful edit. A shared source of knowledge needs people who know the subject and know how to check claims about it.
Sites that earn money need these ties too. A reader can become a specialist writer, a source, a subscriber or the person who points out a serious error. Traffic counts tell us little about which of these roles a reader might take on. Removing the visit does not tell us how many disappear, but it should make us ask how those relationships will form instead.
AI can help people keep producing
Google’s Markham Erickson offered the strongest answer to this concern in June 2026: Google depends on healthy sources of information. Its audience tools, links and paid partnerships can support the sources that make its own services valuable. Fresh facts give an answer service a business reason to invest in the web.
Platforms have good reason to pay for information they need. Support could take several forms:
- Payments could support reporting.
- Links could bring readers who return.
- New tools could reduce the time a small team spends on work that keeps it away from its subject.
Wikimedia has proposed tools to help people do the work. In April 2025, Chris Albon and Leila Zia set out its plan to use AI to help moderate and translate content. It would also help newcomers get started and help people deal with technical problems. People would still make the judgments and set the rules. Help with those tasks could make the work less tiring and bring more people into it.
The outcome depends on what gets easier. Helping a contributor check a source or use an editing system can improve the work. Making endless versions of an answer mostly changes where people can read it. Both can be useful, but they do different jobs.
Nor does a platform’s interest in healthy sources guarantee that support reaches every kind of source. A large business partner is easier to find and make a deal with than scattered experts. Some useful findings matter to few people. Others gain value only when an obscure subject becomes urgent.
We cannot assume those gaps will remain unfilled. Nor can a few deals prove that the whole problem has been solved. If platforms need good information, they should explain how they will help sources keep it current and cover a wide range of subjects.
Pay for the work that comes next
Some publishers are testing ways to fund that work. In Digiday’s September 28 account of its publishing summit, Puck co-founder Liz Goff described tying journalists’ pay to the subscribers they bring in and keep. That gives writers a direct stake in keeping readers. It will not suit every site, but it makes the incentive clear.
A site funded by members will need a different deal from one that sells ads or asks people to volunteer. In each case, can people keep doing the work and checking that it is still right? A count of generated pages or news of a licensing deal cannot answer that question by itself.
Publishers can judge a new deal by what it lets them pay writers to do and keep up to date. Answer platforms should commit to sources for the long term and check whether their support helps the work continue. Readers can support a source they value even when its facts reach them somewhere else.
The harm is a risk we can prevent. AI may help more people take part and improve the work. New ways to fund it may make some sources stronger than they were. But the ability to produce an answer does not make any of that happen on its own.
Somewhere in the material we rely on is an entry that needs correcting and a subject nobody has covered well. Keeping that work possible means giving experienced contributors a reason to stay and newcomers a way to begin. A web rich in answers still needs places where people learn to make the sources.
