.Pizza Glue and Rock Eaters: Google’s AI Overview

Looking for real news and information? Here’s how to not find it.

Search for anything on Google right now, and you may find, at the top of your results page, an AI-generated paragraph that may or may not be useful or accurate, and will definitely not be authoritative information you can trust.

The so-called AI Overview, branded with a tiny, twinkly logo, contains sentences mined (and reworded) from websites that relate to your query. In theory, this feature saves you the effort of clicking on a link to get the information you are looking for.

Liz Reid, Google VP and head of Search, announced the launch of the Overview with a May 14 blogpost cheerfully headlined “Generative Al in Search: Let Google do the searching for you,” and containing a promise to eliminate the last bit of friction for those whose search verb is Google. “With expanded Al Overviews … our custom Gemini model can take the legwork out of searching.”

So: It’s clear that the AI Overview is specifically designed to vastly diminish the need to visit other websites. But what the blog post did not say is: “Search? No need to search. We have the information you want right here. Because we stole it.”

In the two months since Google unveiled its AI Overview feature during the I/O Conference at Shoreline Amphitheater, this purported breakthrough in search technology has been hugely controversial—and borderline disastrous. For one thing, the vaunted Gemini AI language model has delivered hallucinatory results that notoriously recommended that users drink urine and eat rocks.

The AI Overview has also drawn hot criticism from the publishers and other content creators who produce the online properties that allowed Google to build its web dominion in the first place. 

The search engine optimization community started noticing the shift six months ago. “I recently saw a drop in my page views for one of my websites, and while analyzing I saw that some of the keywords where I was ranking in top 3 is not getting anymore clicks,” one SEO specialist posted to Reddit. “Google’s generative AI is scraping the content directly from my website and giving it to the users.”

This phenomenon has been particularly true for the creators that produce the most valuable and expensive content on the web: real and trustworthy news. That sparked concerns that, having used news content to build its vast empire, Google was now making a play to become its own stand-alone news channel.

Ever since AI was first applied to search, by companies including Perplexity and Arc, critics have decried the way these “answer engines” scrape information from sources on the web without directing users to those sources.

Google’s Reid vowed to forgo that practice and predicted that the AI Overview will actually help content creators. “We see that the links included in AI Overviews get more clicks than if the page had appeared as a traditional web listing for that query,” she writes. “As we expand this experience, we’ll continue to focus on sending valuable traffic to publishers and creators.”

Or not.

An investigation released in June by the British media industry journal Press Gazette, performed following Google’s May update, found that publishers saw an immediate decline in organic web traffic.

Headlined “‘Devastating’ potential impact of Google AI Overviews on publisher visibility revealed,” the Press Gazette investigation showed—with extensive use of real-time screenshots—that AI Overviews on news-related Google queries were displacing links to news sources, and did not contain linked attribution.

“Google insists that links contained in AI Overviews generate more clicks, but has declined to share any data on this,” PG editor Dominic Ponsford writes. “The research published today suggests that in many cases the AI Overviews contain no links at all.”

As with most everything on Google, the user experience might change on any day and might be different for every user. On Monday of last week, Google was not sending me any AI Overview results at all, while a colleague on a Slack call was seeing them everywhere. Yesterday, I got Overviews with no links to sources. Today, every search that yields an Overview comes with a stack of sources, including photos.

The Press Gazette’s investigation was systematic, and involved searches in the US for “3,300 of the most important search terms for leading publications.” It found that AI summaries appeared as the top result in nearly 25 percent of cases. In each of those cases, the number one actual web search result, which previously would have gone to a news publisher, was pushed down by a full page scroll.

The Press Gazette calculates that publishers whose content previously would have ranked in the top results spot could see an 85% drop in the number of clicks they receive in search results replaced with an Overview.

“The research found that Google was returning AI-generated summaries for even very newsy searches, effectively setting itself up as a publisher in its own right.”

Google Takes on News 

News media advocates in Great Britain and around the world have reason to be suspicious of Google because of something happening in California right now. This past spring, some publishers here noticed their organic search traffic plummeting. 

To help explain what happened, I conducted the following search: “Did Google remove links to California publishers’ news sites?” The AI Overview delivered this explanation: 

“In April 2024, Google began testing a process to remove links to California news websites from its search results for some users in response to the California Journalism Preservation Act (CJPA). The CJPA, introduced in March 2023, would require tech companies like Google to pay news outlets a ‘journalism usage fee’ when they use their content alongside digital ads.” (Google calls it a “link tax.”) 

Google says the bill would harm its business model, limit the services it can offer Californians, and reduce traffic to California publishers. “We believe CJPA undermines news in California,” Jaffer Zaidi, Google’s VP of Global News Partnerships, wrote in an April blog post. He says the search giant has been “helping journalists and news publishers evolve in response to the rapidly changing way people are looking for and consuming information” but that “people are getting news from sources like short-form video, topical newsletters, social media and curated podcasts.”

One finds a counter argument further down the search results page after following a link to an article on Governing magazine’s web page. Chris Argentieri, president and chief operating officer of the Los Angeles Times, is quoted addressing a hearing in Sacramento.

“Just to understand the difference in market dynamics,” Argentieri says, “just consider that Google earns enough advertising revenue to pay for the [annual] cost of our newsroom in less than three hours. Google’s revenue for a month or two would cover the cost of all working journalists in California.” 

And on CNN we learn that California State Senate President Pro Tempore Mike McGuire, a co-author of the CJPA, called the blocking of news sites an act of “bullying” and an “abuse of power.”

“This is a dangerous threat by Google that not only sets a terrible precedent here in America, but puts public safety at risk for Californians who depend on the news to keep us informed of life-threatening emergencies and local public safety incidents,” McGuire tweeted. “This is a breach of public trust and we call on Google Executives to answer for this stunt.”

McGuire and his Senate colleagues last month sent a bill to the Assembly that, like CJPA, is designed to help the state’s newsrooms. Sen. Steve Glazer’s SB 1327 imposes a 7.25 percent tax on money derived from “data extraction transactions,” defined as practices in which a company “sells user information or access to users to advertisers,” or “provides services to a user in exchange for displaying advertisements or collecting data.”

Photo collage of a person sitting in front of a giant computer monitor
SEARCH TERMS If Google hoards traffic by answering questions with AI, Nilay Patel writes, it could ‘potentially even destroy the internet ecosystem as we know it.’ Photo: Collagery Shutterstock

That legal language, you may notice, pretty much describes Google’s business model. And to make it extra clear who it targets, the tax only applies to “those persons generating $2.5 billion annually in gross receipts derived from data extraction.”

While the fate of these two California laws, and the violence of Google’s response, are uncertain, the AI Overview is here, and its damage is being felt globally.

Owen Meredith, CEO of the UK’s News Media Association, said the AI summary violates Google’s stated mission to “organize the world’s information and make it universally accessible” by sending visitors to websites.

“Introducing [generative AI] into search, and AI Overviews that directly synthesize and present information to the user, risks discouraging users from clicking through to the original links, in turn threatening the business model of those who invest in journalism and quality information,” Meredith said, noting that these changes have been developed by Google “without consultation, transparency, permission, or reward for the original curators and rightsholders of that information.”

The ‘Google Zero’ Apocalypse

This exact scenario had been predicted by Nilay Patel, editor of The Verge (and former copyright attorney), who branded the phenomenon “Google Zero.” He defined his term as “that moment when Google Search simply stops sending traffic outside of its search engine to third-party websites.”

For many months now, Patel has been warning his fellow website publishers that this would be a hugely problematic moment because Google essentially owns the online universe. “The entire web is Google’s platform,” Patel says, “and creators on the web are often building their entire businesses on that platform.”

In an interview with Alphabet/Google CEO Sundar Pichai on the day following the launch of AI Overview, Patel pointed out that the search giant seemed to be abandoning its core practice of sending traffic to websites, where their owners can monetize that traffic.

“Google is by far the biggest source of traffic on the web today,” Patel writes in an intro to his Decoder podcast with Pichai, “so if it starts keeping that traffic for itself by answering questions with AI, that will change or potentially even destroy the internet ecosystem as we know it.”

On the pod, Patel confronted Pichai about the tension between Google’s optimistic vision of AI’s bright future “and the very real fears and anxieties creators and website owners are feeling right now about how search has changed and how AI might swallow the internet forever.”

Donning his old copyright-lawyer hat, Patel argued that the old search model, in which Google took information from websites and displayed that information on its results page in return for links to the content-creators’ websites, would be construed legally as “fair use.” Displaying that information without the promise of a click, he said, is what lawyers call a “taking,” and the rest of us call theft.

Pichai insisted that he and his company have website creators’ interests front of mind. He described a year-long “journey” during which Google Labs iterated what was then called the Search Generative Experience, and said, “I constantly found us prioritizing approaches that would send more traffic while meeting user expectations.

“To the extent there is value there, we obviously think there is a case for fair use in the context of beneficial, transformative use. I’m not going to argue that with you given your background.”

Patel, who has for months been detailing the harms AI search is doing to website owners, pressed Pichai, pointing out that while Google frequently responds to complaints from big corporations, they do not appear to be listening to folks who are mad about the new search ecosystem.

Pichai appeared to be stung by the charge and replied in an almost pleading tone.

“When you look at how we have approached the Search Generative Experience, even through a moment like this, the time we have taken to test, iterate, and prioritize approaches, and the way we’ve done it over the years, I would say I definitely disagree with the notion we don’t listen. We care deeply; we listen.”

As we will see, there is evidence that Pichai was not just spewing pablum.

The Google Whisperer

Danny Sullivan spends his days (and frequently nights) interfacing online with the nerds who care about search, who are known by acronym as SEOs. Most of this happens on social media and web forums: He’s @Google SearchLiaison on X; searchliaison on Mastodon; googlesearchliaison on Threads and searchliaison.bsky.social on BlueSky.

A longtime search guru, Sullivan in fact helped popularize the phrase Search Engine Optimization. He founded the industry bible Search Engine Watch way back in 1997, and a decade later founded Search Engine Land and its Search Marketing Expo (SMX). When Google hired him as the public face of its search efforts, it made a big splash. And not universally in a good way.

Last November, The Verge’s Amanda Chicago Lewis included Sullivan in an article called “The people who ruined the internet.”

“In 2009, Sullivan was described as ‘the closest approximation to an umpire in the search world,’ so when he published ‘A deep look at Google’s biggest-ever search quality crisis’ in 2017 and then took a job as Google’s public liaison for Search only a few months later, it felt to some SEOs as though a congressperson working on gun safety legislation had quit to become an NRA lobbyist.”

Lewis clearly chose the gun metaphor on purpose. Her piece involves a Hunter-Thompsonesque visit to an online marketers’ convention in Florida, and a deep-dive into the various ways the practitioners of the dark art of Search Engine Optimization have weaponized the practice. Throughout, she returns to an imagined scenario in which a 10-foot alligator mauls a digital marketer.

It’s “like the Wizard of Oz projecting his voice to magnify his authority,” Lewis writes of her prey, before stating a simple fact about an SEO’s mission: “The goal is to tell the algorithm whatever it needs to hear for a site to appear as high up as possible in search results, leveraging Google’s supposed objectivity to lure people in and then, usually, show them some kind of advertising.”

Thus, SEOs have turned the World Wide Web into a place where it’s naive to Google anything and expect to find reliable and relevant results.

Sullivan responded to the piece with a 6,000-word blogpost. (!) While he does not take particular issue with Lewis’s analysis of the state of search, he (exhaustively) explains that Google is working diligently to fix the problem.

“The problem is Google’s, and I did not say it wasn’t,” Sullivan writes. “I almost certainly said the opposite, that the fail moments are indeed Google’s. That the expectations are Google’s problems and, importantly, good problems for us to have. Because it means people care about Google Search. They care enough to complain, they want it to improve, and it’s on us to make it even better.

“Google Search is privileged to have so many people trust in it to help them locate information. I’m privileged to be part of that. I fully recognize the importance, the challenge, the trust, and I hope to keep living up to those expectations, humbly so.”

This language calls to mind the “Don’t be evil” era, and Sullivan can be found all over the web delivering the same message—often echoing the pleading tone that Sundar Pichai used when explaining to Nilay Patel that Google cares.

Alien Spaceship Crashing in Palo Alto Building. Photo by Scott Beale / Laughing Squid

“I’ve spent a huge amount of time looking at the feedback over the past few months,” Sullivan told Search Engine Roundtable, “diving deep into sites, writing up thoughts and talking with people internally.

“I fully recognize that [there are] sites that are diligently producing great content but our systems aren’t recognizing it as well as they should … also a lot of really poor content that our systems are indeed recognizing. And there’s also great content that is doing very well—but those doing very well with great stuff tend not to talk about that on the socials.”

Elsewhere, addressing the shameful situation in which searches that include the actual name of a specific website do not yield that website as a result, Sullivan vowed to talk to his colleagues.

“If someone is seeking a particular site by adding the site’s name, we should generally show content from that site (even if it’s not a formal site: query). We should be doing a better job in some of these cases; I’ll flag this to the ranking team.”

Yes—Danny Sullivan’s job is mollifying the concerns of his cohort; and yes this all sure does sound like pablum. But on one key feature—the thing that allows users to defeat the AI Overview itself—he appears to have delivered.

Google Hacks its Own AI

As of this writing, there is a tweet pinned to the top of Sullivan’s personal profile on X. It promotes a major development that leaked out quietly a week before the thunderous AI Overview launch.

“Since I joined Google, I’ve just been a boy standing in front of the search group asking it to love a Web filter. So happy to see it’s arrived—congrats to the hard-working team on this project that through their own efforts made it a reality!”

This is attached to a retweet from his Google SearchLiaison profile: “We’ve launched a new ‘Web’ filter that shows only text-based links, just like you might filter to show other types of results, such as images or videos. The filter appears on the top of the results page alongside other filters or as part of the ‘More’ option, rolling out today and tomorrow globally….”

Click on the Web filter and you will find yourself back in 2007. No “People also ask …” queries to keep you on Google. No links to Google Maps. No Quora links, No AI Overview. The Web search filter takes you to a page with ten blue links to third-party websites.

ArsTechnica was one of a handful of outlets that picked up the Web search news, and published it under a semi-snarky headline: “Google Search adds a ‘web’ filter, because it is no longer focused on web results.”

Perhaps because the piece was written in the wake of the AI Overview shitstorm, the news about the Web search link is buried under some complaints, and concludes with a complaint.

“What you can do is go find a new ‘Web’ filter, which can live alongside the usual filters like ‘Videos,’ ‘Images,’ ‘Maps,’ and ‘Shopping.’ That’s right, a ‘Web’ filter for what used to be a web search engine. Google says the Web filter can appear in the main tab bar depending on the query (when would a web filter not be appropriate?), but I’ve only ever seen it buried deep in the ‘More’ section.”

In other words: To escape from the tyranny of the AI Overview, you will have to execute some clicks. And as Google knows, clicking is the sound of friction.

Will the humble Web link, nestled in the More subnav, prevent Google’s AI from destroying the the internet, the news industry, and countless online businesses? Not likely. And the problem may be even bigger than that.

Arati Prabhakar, the director of the White House Office of Science and Technology Policy—a cabinet-level position, is of course focusing a lot of her efforts on AI, and what she calls its “promise and peril.” In a conversation with Nilay Patel last week, she pointed out the deadly serious paradox that AI presents.

The powerful biodesign tools that are being created to aid President Biden’s cancer moonshot can also be used to create biological weapons. The AI image generators that put powerful creative tools in the hands of millions of people are being deployed to make deepfake porn that is harming countless women and girls. The danger presented by AI is not just something to fear in the future; it is, as she says, “a today problem.”

Among the harms that Prabhakar notes is what she calls “the deterioration of our information environment.” I do not believe she has weighed in on the proposals being put forward by California lawmakers. I do know she is aware of the scope of the problem they are attempting to confront.

LEAVE A REPLY

Please enter your comment!
Please enter your name here

Latest Giveaways

Enter for a chance to win a $50 gift certificate to Capers Eat & Drink in Campbell. Drawing September 25, 2024.

$25 to Nomikai

Enter for a chance to win a $25 gift certificate to Nomikai in San Jose. Drawing October 9, 2024.
spot_img
10,828FansLike
8,305FollowersFollow
Metro Silicon Valley E-edition Metro Silicon Valley E-edition