Traffic Apocalypse – Columbia Journalism Review

Sign up for The Media Today, CJR’s daily newsletter.

Worldwide search traffic has fallen by 15 percent in the past year, estimates Similarweb, a digital intelligence platform. The culprit: AI search. Now that AI-generated summaries are being integrated into search results, anyone looking for information has less reason to click through to the websites where that information originates. For media publishers whose business models rely on referral traffic to bring them advertising revenue, this shift feels nothing short of catastrophic.

There’s no getting around the decline in traffic. Last week, the Pew Research Center released a report showing not only that people who see an AI-generated summary on Google search are significantly less likely to click on external links than users who don’t, but that people almost never click on the links included in the AI summary. (They do so just 1 percent of the time.) Similarweb reported that, in the year after the launch of Google’s AI Overviews, in May 2024, the proportion of news searches in Google where people didn’t click on a single link rose from 56 percent to nearly 69.

Earlier this month, the Independent Publishers Alliance submitted an antitrust complaint against Google to the European Commission, alleging that AI Overviews “have caused, and continue to cause, significant harm to publishers, including news publishers in the form of traffic, readership and revenue loss.” They pointed out that most publishers cannot opt out of having their content ingested for AI Overviews because that would mean exclusion from Google search. 

Google has been rapidly testing and deploying additional AI features across its search ecosystem. In May, the company introduced “AI Mode”—a chatbot-like interface, akin to ChatGPT or Perplexity, that provides synthesized answers to queries with some source links embedded. The release of AI Mode prompted the News Media Alliance to issue a warning that the feature would “further depriv[e] publishers of original content both traffic and revenue.” Other Google experiments have included Audio Overviews, which generate spoken conversational summaries, and Search Live, a voice-based search interface that allows users to engage with the search tool through a back-and-forth conversation. Just last week, Google rolled out a new feature called Web Guide that uses AI to organize search results, producing an output that appears to be a hybrid of traditional and AI-driven search. And Google Discover—the content feed on Android and iOS that offered a glimmer of hope for publishers amid the AI-driven shake-up—has begun integrating AI summaries into its recommendations as well.

All of these features are a result of Google’s scramble to remain relevant. As I wrote a few months ago, rapid innovations in generative AI have intensified competition between tech companies, threatening Google’s long-standing dominance in search and browsing. OpenAI and Perplexity are moving aggressively into the browser market: Perplexity recently launched its Comet browser, and OpenAI is close to releasing its own AI-powered Web browser, Reuters reported. OpenAI is also developing tools intended to rival Google Workspace and Microsoft Office. 

Publishers are suffering the collateral damage of this competition. But Google can’t abandon them entirely. Its products, like those of most AI companies, depend on access to timely and high-quality information. A Muck Rack analysis found that when users submitted queries about recent information, generative AI models responded by citing journalistic content 49 percent of the time. Now, after years of avoiding licensing agreements with news outlets—Google has only one publicly known deal, with the Associated Press—the company is seeking to recruit about twenty news outlets to license their content for an unspecified AI-related project. Additionally, earlier this month, Google partnered with a few media publishers, including The Economist and The Atlantic, to feature their content in its AI-powered research tool, NotebookLM.

To help publishers drive up monetization of their content, last month Google released Offerwall, which lets publishers give their audiences options on how to pay for content. Those options include paying a small fee per article, watching ads, and completing surveys. Google says it tested the feature with more than a thousand publishers, and that smaller publications in particular would benefit, but the company did not provide any data on Offerwall being used successfully. As Sarah Perez of TechCrunch pointed out, many of Offerwall’s solutions, such as paying per article, have already been tried—unsuccessfully.

Many publishers see the traffic declines caused by AI Overviews as a signal to shed their long-standing dependence on Google. “Google is shifting from being a search engine to an answer engine,” Nicholas Thompson, CEO of The Atlantic, told the Wall Street Journal last month. “We have to develop new strategies.” One strategy is trying to demand compensation from AI companies for crawling their content. Media companies are also investing in their own channels that deliver content directly to readers. They are launching new subscriptions, newsletters, events, membership programs, and even platforms and apps.

Wired’s Katie Drummond wrote recently that the solution to this “traffic apocalypse” is not so complicated. The trick, she wrote, is to “connect our humans to all of you humans.” She’s betting that audiences care about the work that goes into creating journalism. Whether they care enough to keep it afloat is another question.