There is a fear right now that AI-centered changes in the web search landscape might end up killing the World Wide Web as we know it. The logic is that websites that publish quality information for a mainstream audience are mostly ad-supported, and if search engines are turning into AI platforms that provide information directly, then there will soon not be a lot of visitors on these websites to view the ads. The irony is that the AI platforms feed on these websites, so without them, there is no "AI search". Therefore, it might ultimately be in the AI giants' own interest to pay for the content they absorb. But I digress.
I can't predict what the web as a whole will look like 5 years from now, but my guess is that commercial information websites will find ways to endure. I might be wrong, of course. What I am absolutely certain of, however, is that the non-commercial web, or small web, will be unaffected.
The term small web was coined to describe a certain revival of non-monetized websites made by individuals, mainly inside geek circles. These sites are sometimes referred to as small websites, which is the term I will use to refer to them for the rest of this blog post. People who make small websites are motivated by a passion, a desire for sharing, or an urge to create or otherwise express themselves. A lot of these websites contain a personal blog (like this one) where the author writes about subjects they are passionate about. There are also purely fun, creative websites to be found on the small web (Floor796 comes to mind). Search engines are usually not the main way these websites are discovered, as people most often end up on them by following links shared by humans on websites like Hacker News. A lot of small websites are not particularly optimized for search engines, and many don't even track views. In fact, views are often not the main motivation factor for people who maintain these websites. What I am trying to say is, whatever happens with search engines and AI, the small web will go on. Actually, I wonder if there is a chance AI could even help it flourish.
The small web has its roots in the early days of the web. Small websites were commonplace back then. Most of them were hosted on services like GeoCities or on web spaces that internet service providers used to provide to their customers. These websites were often ugly by modern web design standards, but they offered a kind of self-expression and creativity that was eventually lost to the rise of modern social media. There are a few projects today dedicated to preserving these now-extinct websites, such as restorativland and Protoweb.
These websites were not just made by "computer people". When web pages consisted only of HTML and images, and when expectations of what a website should look like were low, it was easier for non-professionals to make them. Virtually anyone could pick up the basics of HTML in one weekend, or even use a WYSIWYG tool like Netscape Composer. It can be argued that the basics of HTML can still be picked up in one weekend, but modern expectations set the bar for learning to make a good website much higher.
Nowadays, people who bother making small websites are usually people who are really interested in computers, often programmers or IT professionals. People who are not that into computers, but might still have made a custom website back in the 90s or 2000s, are more likely to post on social media or a publishing platform like Medium or Substack these days. What I am wondering is if AI could bring more of these people to create actual websites? After all, AI lowers the barrier of entry for creating things. Sure, no-code tools pre-date LLMs, but those are more limited, and besides, it's hard to beat natural language in terms of learning curve.
With enough will and agency, and a bit of computer literacy, someone with little to no programming skills can now use AI to create a website, and can be helped by AI to get it online. They can also iterate to make the website look how they want, and can even have the AI change the design or layout in a matter of minutes, thus lowering the cost of experimentation. In other words, LLMs can empower creative individuals in ways that previous tools couldn't. Don't get me wrong: I am not saying "vibe-coders" will replace programmers in the grand scheme of things. In fact, I believe there is now ample evidence that this is not going to happen. And I certainly prefer to live in a world where software is made by people who know what they are doing. That being said, not all codebases matter equally, and a small website being made by someone who does not know what they are doing is rather inconsequential. Pretending non-programmers should not use AI to create small websites would not only be unnecessary gatekeeping, it would also be contrary to the spirit of the early web. A lot of people were creating websites without really knowing what they were doing back then, either using WYSIWYG, or stitching together code snippets from various corners of the Internet. How many web developers of today wrote their first HTML on Angelfire in the 2000s?
Of course, the social media problem remains. Why bother creating a website when it's so easy to share content on a platform instead? My hope is that with the barrier of entry being lower, more people might find it "worth it" to create a web space that they actually control and can be more creative with. I myself have experienced the new "worthiness" of making things myself with the help of AI. These days, I find that I am much more likely to make my own "home-cooked tools" when existing apps don't behave exactly like I want. In a pre-LLM world, I would have settled for an existing app that I was not fully happy with. With LLMs, I am less likely to settle. I must not be the only one.
Maybe LLMs can even get some people into actual programming. I firmly believe that curiosity is the most important quality for working with computers. And it takes some amount of curiosity for someone with no programming experience to have the idea to use an LLM to make a website, and to act on it. The most curious will eventually look at the code, and maybe try to figure out how it works. The next generation of teenage computer nerds might very well be using AI to learn about software development right now.
Finally, LLMs can help small website ideas come to life even for people who do know how to code. For example, in the spirit of "fun and creative" websites I mentioned earlier, I had dreamed for years of adding a "retro mode" to this website, which would simulate the look and feel of the 90s web. I would never have had time to implement this "by hand". I finally did it with a coding agent (look for the 90s Simulator link on my home page). Of course, the code is a mess, but again, that doesn't really matter for a small website (of which the codebase was already a mess anyway: there is a reason it's not open source, and this reason is shame). It was still a creative endeavour too: in order to do this, I spent a lot of time browsing old websites on Netscape Communicator using ProtoWeb's proxy that simulates a dial-up connection. I had fun making this, even if I barely wrote any code. And fun is an important part of the small web for me. There are a lot of bad things about AI, but if AI can help make the internet more fun, that's at least one good thing.