You are reading a State of Dystopia post. These entries deal with current events that put us on the cyberpunk dystopia timeline. Read them now to see the future we’re going towards. Or read them in the future to figure out where things went wrong.

I see decline everywhere. In civil liberties, in democratic norms, in the economy. In and of the entire country. Yet I took for granted that we had figured out the art of searching for information, so the decline of Google Search has been shocking to me – akin to suddenly appreciating the fragility of modern life during a blackout.

If the decline of Google search has been noticeable for years, it became inescapable this year. So many results are irrelevant to the search term. Both the relevant and irrelevant results are increasingly low-quality – designed to target keywords, and to compete with every other high-ranking result. Fact-checking is much more circuitous, and complex queries are barely worth trying anymore.

AI is often described as the first real threat to Google. I have long been skeptical of that claim. Until about a couple weeks ago, when I tried using to fact-check something for work, and found it more helpful than Google.

Since then I’ve been using Perplexity about as often as I have Google. I currently have just 2 use cases: asking Perplexity to fact-check short statements, and asking it to generate a list of options (essentially a stand-in search results page).

My own shift to Perplexity is an example of the supposed threat facing Google, and part of why Google has rushed to incorporate generative AI into its consumer-facing tools.

But this shift is not what the so-called innovators describe it as. It’s not that generative AI is a fantastic new product for search – it’s that Google has gotten much worse. Perplexity is a reasonable substitute for now, but it’s not even close to being as useful as Google was 10 years ago.

Perhaps ironically, AI is at least partly to blame for Google Search’s decline. ChatGPT and similar tools don’t produce great work, but they can certainly produce a lot of passable work, with the result being that websites can crank out tons of crap that fills up search results. This was a problem before the mainstream advent of generative AI, but it has ballooned since then.

We’ve all seen the screenshots of Google’s AI Overviews recommending users put glue on pizza or eat rocks. (My personal suspicion is that Google does not trust most of its own search results, and so privileges answers on forums, but there’s no evidence for this and there’s a more immediate technical explanation anyway.) But these results comprise just a small slice of searches, and the feature will likely get some improvements.

So while I hope Overviews is not here to stay, it probably is. And with it, the odds that Google will ever improve its deteriorating search function get even slimmer.

The primary reason this is all happening, of course, is the obvious one: Google is a monopoly. If its search tool goes to shit, it truly does not matter in the eyes of Google: it has quashed all other would-be competitors, and aside from the massive infrastructure advantage it has over any potential upstarts, it can pay as much as it needs to ensure every device and browser uses Google Search by default.

And it’s not that Google Search is an unimportant source of income. Yes, Google has half the mobile OS market and a robust cloud hosting business. But approximately half or more of its revenue comes from Search. From Alphabet’s 2024 Q1 results:

Note: ‘Google Search and other’ refers to Search as well as Google properties like Gmail, Google Maps, Google Play, etc. Altogether, it accounts for 57% of Google’s revenue. Meanwhile ‘Google Network’ includes revenue from AdSense, which of course is heavily embedded in Search.

Google does not care about the quality of its most important product because quality became disentangled from profit some time ago. The first quarter report above showed a 15% jump in advertising revenue from the previous year, much of it from Search. Which is why Google happily authorized a $70 billion stock buyback even as its flagship tool withers.

Since the 1980s, US antitrust enforcement has revolved around whether a merger/company will harm consumers, rather than the question of whether the company attained market power by anticompetitive means. Until the recent tech-oriented antitrust push, this made it easy for tech companies like Google to argue that its overwhelming market power benefits consumers.

And while tech hawks (like yours truly) have plenty of good counterarguments, it’s hard not to begrudgingly concede that Google has provided plenty of benefits to consumers. After all, Google Search is free. Gmail is free. Chrome is free. And for anything to compete with Google, it has to be free too (which is indeed a problem, but not under the consumer welfare standard described above). Seems quite distant from the days of Standard Oil.

But now the cracks in that defense are widening by the day. The federal government’s ongoing antitrust case against Google alleges, among other things, that Google’s monopoly in Search has resulted in a reduction in quality for the typical user. Google’s legal defense could still win out, but it’s a point for cautious optimism.

But getting recognition of Search’s decline is just a small victory, and even the ultimate victory — a breakup of Alphabet — would not guarantee that Google Search returns to its peak. The current trend away from quality search is not necessarily permanent, but it has plenty of momentum. And if it holds, the ramifications extend beyond the nuisance of seeing spammy websites.

Everyone – centrist liberals, neo-Nazis, Marxists, whatever – seemingly agrees that American society is being dumbed down. It’s an assertion I’d caveat endlessly, but at least as far as the last 10-15 years go, it does seem modern tech has had a dulling effect. It’s certainly destroyed attention spans. (Whether it’s made people less capable of critical thinking is not so clear yet, to me anyway.)

Until now, threats to our cognitive ability came from other areas of consumer tech: social media, video-sharing platforms, smartphone interfaces.

The twilight of the search engine is a quiet continuation of that pattern. You’re not supposed to ‘search’ anymore — you’re just supposed to passively receive answers. Which is why Perplexity calls itself an ‘answer engine.’

Yes, finding answers with the Google Search of 5 or 10 years ago had more slightly friction than finding answers with Perplexity does today. But only because users had to read a thing or two, and had to consider what to even look for. Search was hardly a rigorous exercise, but it did require something — and now even that looks ready to slip away. Perhaps the ‘answer engines’ of the near future will still require some thought in prompting, but one suspects that as they improve, they will preemptively answer more, and render the user’s role as passive as possible.

Imagine the brainrot associated with social media addiction, and extend it further. Someone who only digests information in the form of short videos with constant animations, who was already too lazy to learn anything by Googling it — such a person now has a tool that’s more their speed. They ask, they get an answer, and they can move on. It’s simple, and that’s the danger.

Part of the marvel of the internet was that it promised to grant everyone, everywhere, instant access to the world’s knowledge. But as one of the most basic functions of the modern internet degrades, will that last?

As things stand now, the more immediate and likely scenario is that a gap will grow between those who have high-quality information indexing and those who don’t. People within certain fields or industries will naturally be best-linked to specialized sources of high-quality information. As accurate information becomes less convenient to access through traditional, consumer-facing tools, these professionals will cling more closely to their existing channels.

Meanwhile, everyone outside of a given industry will be stuck with an increasingly shitty Google Search and/or the answer engines. Typical internet users will still be able to get most of the information they want, but high-quality information would be harder to come by, increasingly siloed within certain professional or even hobbyist communities. These high-quality repositories could simply take the mundane form of gated forums or Discord servers.

This is even more likely if a group feels threatened by automation. Think here of lawyers or journalists. Both of these groups of workers will feel the pressure of AI bearing down on their jobs, if they don’t already, but both groups are generally well-organized and influential. They might succeed in securing some level of job security, and part of that security will include their relationship to information. So they might restrict the use of industry content in LLM training sets; create or buy access to higher-quality search engines; create internal chatbots or ‘answer engines’ trained on industry textbooks and manuals; and probably other things I can’t think of.

This is highly speculative and should obviously be taken with a grain of salt, but the seeds of it may already be here:

In late May, The Atlantic and Vox Media signed licensing deals with OpenAI. OpenAI will get to train on archived content, and therefore improve the general credibility of its models without risking copyright infringement. The media outlets, meanwhile, get compensated — and cemented as authoritative sources, because ChatGPT will be obligated to cite and link to them. But even as journalists get approved as a source of information for everyone else, will they themselves rely on ChatGPT to obtain information? Beyond scanning lengthy documents, I doubt it.

To be clear, having a learned upper class is hardly a new phenomenon. Aside from being the general story of civilization, it’s been a prominent theme in our own society for the last few decades, with the professionalization of all sorts of jobs. (To again use journalism as an example: master’s degrees are not required for most journalism jobs, but are incredibly and increasingly common anyway.) So it’s not that this scenario would mark a truly new development, it would just cement a broadly negative trend.

Perhaps a small segment of everyday users will switch to premium search engine upstarts. Popular tech commentator Cory Doctorow, in a recent blog post about Google’s worsening quality, described astonishment when first using a new search engine called Kagi:

I tried it. It was magic.

No, seriously. All those things Google couldn’t find anymore? Top of the search pile. Queries that generated pages of spam in Google results? Fucking pristine on Kagi — the right answers, over and over again.

But the catch is that Kagi is a premium service. You have to pay for the quality. And maybe that’s the way it’ll just have to go from now on: an enlightened few will start paying to have decent search engines, whether out of professional necessity or personal preference, while everyone else will stick to a reduced Google Search. (And maybe answer engines.)

Some would say it’s a long overdue development: that our universal reliance on free online services is harmful and that it’s about time we just started forking over money for good products. But as a long as a free and easy default exists, with a 20-year cultural presence and overwhelming market power, I doubt the paid models will take off en masse. It’s a debate for another day, but for now, suffice it to say that the era of high-quality and free information indexing seems to be ending.

The death of the search engine is not yet a certainty, and if it happens, it just heralds a return to the historical norm. But that would be sad in its own right; while the internet never reduced inequality materially, it did at least level the playing field for anyone seeking knowledge. Information has always been a privileged commodity, but for a time it was much more open. It was progress, while it lasted.

Get more like this in your inbox. It’s free.

Make a one-time donation, if you’d like. (I’ll still put out posts even if you don’t.)

Choose an amount


Or enter a custom amount


Your contribution is appreciated.


*This post was updated on July 2, only to remedy small errors and duplicate text. The content is otherwise unchanged.

Leave a Reply

Your email address will not be published. Required fields are marked *