/media/cinemagraph/2023/07/11/google-animate.mp4)
There’s no such thing as quickly finding a recipe online. Every cooking blog, it seems, is stuffed with paragraphs and paragraphs before you get to an actual recipe, text that might tell you about the history of the dish, or even which of the author’s kids likes it best. It’s not that food writers are categorically long-winded, though. They are just responding to Google’s demands.
The company’s search algorithm appears to favor original text over a recipe you can find in many places online, so food blogs try to game traffic by adding more text. For the past two decades, even subtle changes to Google’s search algorithm have steered not only what content shows up in search results but what kind of content people actually make. “There are loops,” Chris Hoffman, until recently the editor of the technology-tutorial website How-to Geek, told me. “Google communicates what it wants. You listen and you survive, or you don’t listen and you die. So you have to follow it.”
But all of the past changes could pale in comparison to what comes next. Google Search, like the rest of the internet, is pivoting to generative AI. The first step is Search Generative Experience, an experimental tool currently available as a public beta. Instead of sending you off to other corners of the web, more search results appear within Google. Sort of like ChatGPT, it pulls information from various websites, rewords it, and puts that text on top of your search results—pushing down any links you see. In the process, it stifles traffic to the rest of the internet, lessening the very incentive to post online. With AI, Google Search might eventually set off a doom loop for the web as we know it.
Google might not feel like it has any choice. The explosion of ChatGPT poses perhaps the biggest threat to the company in its history. Instead of poking around on Google for a recipe and then scrolling endlessly, now you can just ask ChatGPT to create one for you (though it might not be particularly good). Search Generative Experience is a trial run that ends in December, and a company spokesperson declined to lay out a timeline for when it might be everywhere. “Any experiences that come to Search will likely look different from what people see,” the spokesperson said in an email, without specifying further. Other services, namely Microsoft Bing and DuckDuckGo, already have widely accessible searchbots.
Google’s bot isn’t like “snippets,” the ubiquitous feature that pulls a bit of text from a website to answer simple questions and then tacks on a link if you want more information. Depending on your search, the bot now grabs information from several sites and compiles whole paragraphs or even lists. For example, type best laptops, and it will fill your entire search page with text, not links. At the top is a list of factors to consider before making your purchase (“Display: Higher resolution equals better picture quality”). Below that is another list, recommending laptops for different scenarios. Apparently the Apple Macbook Air is “good for working from home,” but some reviews note that “the camera notch encroaches on the display panel.”
It is all convenient, but where is the information coming from? There are a few links in the top-right corner, or you can click a tiny button to find a rundown of the sites that the bot used to create the result. (The list of laptops Google fed me was generated using information from articles in Engadget, USA Today, and PCMag, among others.) But chances are, most people won’t click any of these links. “When you take all that content and slap it on the search page, people don’t click through,” Hoffman said. That certainly has been true in my time testing the service. Some searches still just give you a normal list of links, but when the results are right there at the top, there is often no point in ever leaving Google.
That could be disastrous for the rest of the internet. Google, after all, dominates the global search market, sending billions of visits to websites across the web every day. All of the articles and product reviews that went into the bot’s guide to buying a laptop presumably thought they would be getting the clicks from searches for best laptop. Much of this sort of content only exists because people are Googling for answers. Hoffman said that while he ran How-to Geek, the site reached more than 30 million people every month; the overwhelming majority of them were searching for tech help. I’ll admit I’m not a neutral party here. I write online tutorials for a living, helping people troubleshoot common tech questions. Millions of people read my articles every week, almost all of them because of Google. After a month of using Google’s searchbot, I’m wondering if I should find another line of work.
If widely implemented, the peril of AI search is that it could eventually shrivel up much of the content that is now reliant on Google traffic for survival. Nothing is going to happen to the internet overnight, but websites hire people like me to notice and solve problems for readers because helpful information gets clicks. If publications aren’t rewarded with traffic, they are going to publish less information. Google claims it understands the stakes, though the company is notoriously secretive about how its algorithms work: The spokesperson told me the company will “continue to prioritize approaches that send valuable traffic to a wide range of creators and support a healthy, open web.” But the basic premise of a searchbot necessarily involves more text in Google, and less traffic to websites. Google’s AI doom loop may lead us into a much smaller version of the internet, with fewer sites, fewer posts—and thus a worse experience for all of us.
The irony is that by cannibalizing the web, Google search results may get less useful. For generative AI to explain something to us, it needs someone to write about it first. Take video-game walkthroughs, in which sites obsessively document every aspect of a game for obsessive fans. AI search, to answer video-game questions reliably, likely requires that these walkthroughs already exist. “You still need someone to actually play the actual game,” V. Buckenham, a video-game developer who has written about AI, told me. The same goes for questions about carpentry, hiking, wine-making, and any other hobby you can imagine—the bots, for now at least, can’t engage in these activities, meaning they need a human to do it for them and write about their experience.
In a world where Google Search is no longer a reliable source of traffic, expect more sites to put up paywalls. Traditionally, sites with paywalls still show up in search results, because they specifically opted in to be there. Will they keep doing that if Google could take their work and feed it into the AI wood chipper? Professional writers, of course, aren’t the only people who publish online. Google could rely on people who post useful information online as a hobby, or as part of an online community, but there’s a hiccup: The age of social media is ending. Group chats on apps such as Discord seem to be growing in popularity, but they are not indexed by Google.
Much of how, exactly, these bots will transform search remains unknown. But at worst, they could move us one step closer to a future in which fewer publications and ordinary people are putting content on the open web. Who will still be publishing? Content marketers who are paid by companies to publish information about their products to influence consumer behavior. Search Google for the best Slack alternatives, for example, and many of the posts you’ll find are written by companies that compete with Slack. According to the job site Zippia, for every one journalist in the U.S., there are already five content-marketing managers. This gets even more complicated when you realize that the content marketers themselves are embracing AI text in a big way, meaning search results could in theory become a bot’s summary of articles written by other bots. Already, Amazon’s best-seller lists have been flooded with AI-generated nonsense.
Google is turning to generative AI for the same reasons as many other companies, but the sheer dominance of Search makes the company’s moves uniquely troubling. “When you’ve got a 90 percent market share, you cannot increase your bottom line by improving your product, by definition,” Cory Doctorow, a prominent tech critic, told me. “If you can’t grow by making the pie bigger, you grow by extracting more from it.” Publishers, who previously got traffic from Google, could get a lot less of it. Users who previously spent a few seconds on Google before clicking a link could now hang out on the search-result page longer. And advertisers, long reliant on Google, could have even fewer websites to work with as bots reduce the traffic to sites that aren’t Google. This is a way for Google to strip-mine the web. But strip mines, by their very nature, aren’t sustainable.