← All Posts

Generative AI is destroying the web

Posted by

  critique tech web

4 min read | 1161 words | 3 views | 0 comments

By this time, most people working in technology or adjacent to it are probably aware of the current frenzy over generative AI technologies, as well as the resistance to it. There are many reasons to be skeptical or even disdainful of these technologies — its impact on the environment has earned it the wrath of environmentalists, and corporate agendas to force AI into every product and service imaginable has earned it the wrath of just about everybody. There are many harms and ills of this technology, and while each one could fill up a post of its own, I'll focus specifically on how it has impacted the web.

Generative AI (or GenAI for short) strives for quantity over quality. It is the epitome of a mass production technology, one that churns out content without respect to quality. This, combined with seemingly every company trying to capitalize on this frenzy (or dare I say fad?) has already backfired in some interesting ways, specifically on the web, which is one of the primary mediums through which users interact with the technology today. Search engines are a good example of this phenomenon. The major search engines — Google and Bing — have integrated AI results into searches, to the extent that they are typically the first thing that users when completing a search. This has already had some far-reaching consequences — it's upset the longstanding symbiotic relationship between websites and search engines that's existed since shortly after the advent of web search. But this is just the beginning (literally).

For those who wish to ignore AI-generated results, these summaries can range from a distraction to a significant annoyance — Bing, in particular, made its Copilot AI summaries so intrusive that it decimated the usefulness of Bing altogether, to the point that I had to completely stop using it. But GenAI's impact on search has been much worse than what search engines themselves have done. Lately (perhaps the past year or so, but maybe longer) I've noticed a lot of web results — articles that appear in search results, that is — which seem to be nothing more than a bunch of GenAI-generated cruft. By this, I mean that it seems as if somebody asked a GenAI tool to write an article about a certain subject, and then published the output as a webpage (personally, I've noticed this especially with HVAC or home improvement related material, though I'm certain it's not limited to these areas).

How can I tell? Well, it's difficult to be sure, but there are some clues. For one, many of these articles are very similar, both in terms of structure as well as content. Often, the articles have inaccuracies, don't articulate the topic well, or state the same obvious points without really addressing any intricacies of the topic. The domains also vary, so it's not clear to me if there's some kind of syndicate publishing AI-generated junk on a variety of websites, or if multiple sites have independently capitalized on this idea. In any case, it doesn't really matter.

These articles are not an asset to the web; in fact, they are the complete opposite — they dilute its value by decreasing the signal-to-noise ratio. It's analagous to the "reverse network effect". The traditional network effect is that as more and more nodes become linked in a network (such as more sites becoming connected to the Internet), the overall usefulness of the network increases. This was originally true with fax machines (hence the related term "fax effect"), and has historically been true of search as well — a search engine that indexes billions of webpages is more useful than one that only indexes millions. But what happens when many of those results are low-quality, likely AI-generated content, which seems to be becoming increasingly prevalent? Rather than helping the user, this is garbage that the user now has to sift through and ignore — in other words, noise. Such results serve no useful purpose, and only serve to reduce the value of web search as a tool. It's quite possible that, after thirty years of mainstream usage, search engines have already peaked and are now on a declining trajectory of decreasing usefulness.

The infiltration of GenAI, combined with the already downhill trajectory of search in general, sometimes referred to as the "enshittification" of search, has gotten a lot of people interested in alternative search engines. Recently, I myself have been turned onto DuckDuckGo Lite, which is an extremely lightweight search engine — just plain HTML, no JavaScript, only a couple web requests, bandwidth-lite, and importantly, free of Generative AI (at least in the search tool itself). I suspect that over time, as more people become more frustrated with GenAI's infiltration of search, more people will seek out engines like this that actually respect the user's priorities.

And while I've been focusing primarily on web search in this article, GenAI's impact is not isolated to that — the trail of destruction extends to much of the web in general. The web is no longer the sea of static HTML documents that Tim Berners-Lee brought into existence. These days, it is an incredibly interactive medium, and also an incredibly fragile one, at that. Because of the prevalance of web crawlers scraping content for GenAI, often without attribution or promise of traffic, as search engines have historically provided, many websites have resorted to primitive measures to block such activity — typically, firing into the dark based on the user agent, catching innocent users in the crossfire. GenAI integrations on sites add even further bloat, much of which users have not asked for and don't want. I wrote previously about how the web's greatest strength — its flexibility — is also its greatest weakness. I very much subscribe to the "function over form" philosophy, and it seems that the web has done a complete 180 from being purely functional (the early HTML-only web, pre-CSS) to a web that is full of form (overly styled and scripted interactive pages) but sorely lacking in function — pages that only work in a few browsers today are merely normal, and truly accessible and cross-platform webpages have almost gone extinct.

This is a very turbulent time for the web, and it's a little scary how much damage Generative AI has done to it in just a couple short years. No doubt, pursuit of profits above all else will continue to drive further "advancements" in GenAI, all while frustrating users, eroding customer experience, and polluting the environment. While not necessarily unique — many so-called "information age advances" have resulted in more automation at the expense of originality and craftsmanship — the rate of change here is even more accelerated. Tech CEOs boast that Generative AI is rapidly transforming society — this is certainly a lofty claim, but there's no doubt it has had a profound (albeit negative) impact on the web. After seeing what's happened in just a few years, I wonder how useful web search and the web in general will still be in ten years. Only time will tell.

← All Posts


Comments

Log in to leave a comment!