What Is a Website For When AI Gives the Answer?

I spend an unreasonable amount of time maintaining this website.

I fix old links, rewrite posts that no longer sound like me, and maintain redirects for paths I changed years ago. None of it is difficult. It is still a lot of work for something many readers may never visit.

That bothers me more than whether Google traffic falls by 10% or 50%. A website used to be the source and the destination. Now it can supply the answer without receiving the reader.

The web is becoming a database that other products query on our behalf.

We Are Already Learning Not to Click

Pew Research Center observed 68,879 Google searches made by 900 U.S. adults in March 2025. When an AI Overview appeared, people clicked a traditional result on 8% of visits. Without one, they clicked on 15%. Only 1% of visits with an AI Overview led to one of its cited sources.

That doesn’t mean every publisher lost half its Google traffic. It does show that once the answer is on Google, leaving becomes optional.

I do this too. I ask ChatGPT or the model in my browser. It searches several pages and gives me something usable. I check citations when the answer looks uncertain or the stakes are high. For an ordinary question, I often don’t.

ChatGPT Search is convenient. It also removes the moment when I might have encountered a writer or an adjacent idea.

Search used to send me somewhere. Answer engines try to make somewhere unnecessary.

Websites Are Becoming Infrastructure for Answers

The old search bargain was easy to understand. A crawler indexed a page and search sent some people back. Publishers turned those visits into subscriptions, sales, reputation, or community.

AI systems consume much more and can return much less.

Cloudflare measured the imbalance across its network. In July 2025, it saw roughly 38,000 Anthropic crawler requests for each referred visit, 1,100 for OpenAI, and 195 for Perplexity. Native apps can omit referral headers, so those ratios may overstate the gap. Even so, this isn’t the old bargain.

A page can now be heavily used while appearing to have no audience. Its facts travel while its author sits behind a citation almost nobody opens.

We assumed Google would build the brain on top of the web. Instead, people may ask ChatGPT, Claude, or Perplexity while the web underneath becomes interchangeable plumbing.

Google Says the Web Is Still Healthy

Google disputes the bleakest version of this story. In August 2025, Google said total organic click volume from Search was relatively stable year over year and that it was sending slightly more “quality clicks”—visits where people did not immediately return to Search.

Google also said people increasingly click first-person sources such as forums, videos, reviews, and original posts. I can believe that. AI can compress “How do I configure this?” It has a harder time replacing “I tried this for six months. Here is what broke.”

I don’t have Google’s aggregate data, and Pew studied one panel of U.S. adults for one month. Both findings can be true. Total clicks can remain stable while informational publishers lose the visits that sustained them. More searches can hide fewer clicks per search.

Websites won’t disappear. Banks, stores, governments, and software still need places where people can act. The vulnerable part is the explanatory web: tutorials, reporting, forums, and personal essays created so another person can understand something.

The Dead Internet Is More Plausible as an Economy Than a Conspiracy

I don’t mean the literal conspiracy. I mean a more boring failure loop:

1. People stop visiting original pages because generated answers are faster.
2. Publishing original work becomes harder to justify or fund.
3. Cheap generated material fills the gap.
4. Answer engines ingest that material and produce more material shaped like it.
5. Actual experience becomes harder to distinguish from an average of previous writing.

The internet does not go offline. It becomes haunted: pages everywhere, authors nowhere.

Readers also lose who is speaking, what they observed, and whether they changed their mind. A synthesized answer can keep the conclusion while stripping away the conditions that made it true.

The web’s messiness was useful. Following a link exposed me to the source’s tone, uncertainty, and neighboring ideas. Answer engines remove that friction. Sometimes that was where trust formed.

So What Is a Personal Website For Now?

I no longer think the best argument for a personal site is traffic.

This site is my source of record. It shows what I wrote, when I changed it, and how one idea connects to another. A social feed doesn’t choose the presentation, and a model doesn’t flatten every paragraph into its house style.

It is also evidence that a person remains on the other side.

Keeping it lively is work. The archive wants to become a dump. Old posts lose context. Generic phrasing creeps in. I have to keep editing, removing, and sometimes admit that something doesn’t deserve to be published.

AI makes that maintenance easier and more necessary at the same time.

I expect fewer people to visit websites for simple answers. The sites worth visiting will offer direct experience, a recognizable point of view, primary evidence, community, or something useful to do. The rest may still be read, mostly by machines.

My response is not to publish more. It is to leave a clearer human trail.

Maybe that is the job of a personal website now: to prove that someone was here.