For years, publishing your thinking on the internet was a pretty good deal.
You spent time researching a topic, forming an opinion, and writing it down. Someone searching for that topic would find your article, read it, and perhaps reach out to you. That could lead to a job interview, a consulting engagement, a conference invite, or a professional relationship that compounded over time.
The return was not necessarily money. It was attention from the right people, and that attention often converted into opportunities.
I think AI is changing this equation.
The internet is increasingly being consumed by systems that can read, summarize, and use information without requiring a human to visit the original source. The person asking the question gets the benefit of the research, but the person who did the original thinking may get nothing in return.
This is a problem of incentives. And if the incentives change, the kind of content people publish will change too.
The reader is changing
Consider what happens when someone wants to understand a new topic today.
They can search Google, open a dozen links, and spend an hour reading. Or they can ask an AI assistant to research the topic, compare different perspectives, and give them a recommendation.
The second option is often faster and more useful.
The AI can read your article, extract the central argument, combine it with information from other sources, and apply it to the user’s problem. The user may never open your website. They may never know who you are.
Your article has still created value. It has helped someone think through a problem or make a decision. But the value has been captured somewhere else.
I have seen some early data on this. A 2025 Pew Research Center analysis found that users clicked a traditional result in 8% of visits when an AI summary appeared, compared with 15% when it did not, and links inside the AI summary received clicks in only 1% of those visits.
This is the important distinction. The problem is not that people have stopped consuming knowledge. It is that consuming knowledge no longer necessarily requires visiting the person or website that produced it.
consuming knowledge no longer necessarily requires visiting the person or website that produced it.
And an agent does not behave like a human reader. A human might read an article, follow the author, share it with colleagues, or remember the person who wrote it. An agent can extract the useful parts and move on.
If I publish an article that helps someone solve a business problem, I would ideally like that person to know who I am. Maybe they read another article, reach out, or think of me when an opportunity comes up. That is part of the return on publishing.
If an AI agent takes the idea and delivers the answer without ever sending that person my way, the intellectual contribution still has value, but I lose much of the opportunity that came with sharing it publicly.
The incentives to share are changing
Companies have understood the economics of information for a long time.
They invest in research, product development, patents, and trade secrets because knowledge can create a competitive advantage. They decide what to share, what to protect, and when to reveal something.
Individuals have historically operated under a different set of incentives. A professional sharing a research paper, a founder writing about a business model, or a product leader explaining how a market works could benefit from making that knowledge public. Publishing was a way to build a reputation without necessarily charging for the content itself.
The internet made this particularly attractive. You could publish once and reach people across the world. Search engines helped the right audience find you, and social networks helped your work spread.
AI changes the distribution mechanism.
The information can still be discovered and used, but the person who created it is no longer guaranteed a relationship with the person who benefits from it.
This does not mean everyone should stop publishing. It means the calculation is changing. The more valuable your thinking is, and the easier it is for an agent to extract that value without sending attention back to you, the more reason you have to reconsider what you share openly.
What happens to the open internet?
I have always found Charlie Munger’s advice about incentives useful: look at the incentives and you can often predict the outcome.
Apply that principle here.
If people get less professional recognition, fewer opportunities, and less revenue from publishing their best work, some of them will publish less of it. Others will move that work to places where access is restricted.
The potential consequence is more interesting than a simple decline in website traffic.
The open internet depends, to some extent, on people being willing to contribute useful information in exchange for something they value. That return might be money, reputation, distribution, customers, or professional opportunities.
If that return weakens, the supply of new, original thinking could weaken as well.
There is a feedback loop here. If more original thinking moves behind closed doors, public sources become less useful for understanding new developments. If the open web becomes less useful, there may be even less incentive to invest in producing new public content.
I do not think this means the internet suddenly becomes useless. There will still be news, entertainment, company announcements, product documentation, and plenty of content produced for commercial reasons. But the supply of thoughtful, independent analysis could change.
The question is whether the open web will continue to be where people share their best ideas, or whether it will increasingly become a place for discovering that those ideas exist.
The best information may move behind closed doors
One possible response is to put more content behind paywalls, paid newsletters, private communities, or membership-based forums.
This already happens for reasons unrelated to AI. Publishers need revenue, experts want to monetize their knowledge, and communities often become more valuable when membership is restricted.
AI could accelerate that shift.
If someone has spent years developing expertise in a particular field, they may decide that publishing their most valuable insights publicly is no longer worth it. They might instead share them with paying subscribers, clients, or a smaller professional network.
There is a trade-off. A paywall can create a direct commercial relationship with the reader, but it also reduces the reach of the content. And restricting access does not automatically prevent AI systems from obtaining information through authorized users, summaries, or other sources.
Still, the underlying incentive is clear: when public distribution stops producing enough value for the creator, controlled distribution becomes more attractive.
Will we start writing for humans and hiding from AI?
There is another possibility.
Today, a lot of effort goes into making content easier for search engines to discover. SEO is built around the idea that if your content is useful and discoverable, people will visit your website.
What happens when the system discovering your content is an AI agent that can take the useful parts without sending anyone to the site?
One response could be to make content easy for humans to read but difficult for machines to extract or use. That might involve technical restrictions, access controls, or deliberately adversarial content designed to interfere with automated systems.
Some versions of these techniques already exist in the broader world of web scraping and AI data protection. But I would not assume that a reliable solution exists for selectively hiding information from AI while preserving the exact same experience for human readers.
There is also a practical problem: if your content becomes difficult for agents to understand, it may become less likely to be recommended by them in the first place.
The more interesting question is whether the internet will develop new ways to distinguish between reading information and using it commercially. We may see more licensing, authenticated access, and systems that allow creators to set conditions on how their work is used.
We are already seeing early examples of this forming. Reddit restricted broad API access and later reached a reported data-licensing agreement with Google worth about $60 million per year, and Cloudflare in July 2025 began asking new domains whether AI crawlers should be allowed and introduced a permission-based model distinguishing training, inference, and search.
The technology is still evolving. The incentive to find a solution is already there.
Easier access doesn’t solve the incentive problem.
I am not arguing that AI is bad for knowledge or that people will stop thinking. AI makes research more accessible, reduces the cost of exploring unfamiliar subjects, and helps people use information that would otherwise take hours to find and understand.
Those are meaningful benefits.
But we should separate two questions.
The first is whether AI makes knowledge easier to consume. I think the answer is clearly yes.
The second is whether the people who produce valuable knowledge will continue to have enough incentive to share it publicly. That answer is less clear.
The internet’s original model worked well because the creator and the consumer were connected. The creator published something, the consumer discovered it, and some of the value flowed back through attention, reputation, relationships, or revenue.
AI weakens that connection. It can preserve the usefulness of the information while removing much of the relationship between the source and the person who benefits.
We are going to have to figure out how that relationship works in an agent-driven world.
As for me, I will keep writing. Writing forces me to organize my thoughts, question my assumptions, and make my reasoning explicit. I get value from that process even if nobody reads the final article.
But I also recognize that the economics of publishing are changing. If we want the open internet to remain a place where people share original thinking, we will need to find ways for the people who create that thinking to benefit from sharing it.
Otherwise, we may end up with AI systems that are very good at answering questions using knowledge that fewer people are willing to publish in the first place.
And that would be a strange outcome for a technology that promises to make knowledge more accessible.
Disclaimer: The opinions expressed in the blog post are my own and do not reflect the view(s) of my employer.