When Frank Pine searched Google for links to news articles two months ago, an artificial intelligence-generated paragraph on the topic appeared at the top of the search results, and he had to scroll through it to see what he wanted to see.
The experience has soured Pyne, editor in chief of MediaNewsGroup and Tribune Publishing, which owns 68 daily newspapers across the U.S., that he now says he dreads reading the passages.
In May, Google announced it would begin rolling out AI-generated summaries of content from news sites and blogs on search topics across the U.S. The change has Pine and many publishing executives worried that the summaries will drastically reduce the amount of traffic Google drives to their sites, posing a major threat to their fragile business models.
“This has the potential to stifle the original creators of content,” Pine said, adding that the AI summaries feel like another step in generative AI replacing “the publishing houses that have been gobbled up.”
In interviews, media executives said Google had put them in a tricky position: They want their sites to appear in Google search results, which for some publishers can account for more than half of their traffic, but doing so means Google can use their content in its AI Overviews summaries.
Publishers can also protect their content from Google by preventing web crawlers from sharing snippets of content from their site, but in that case the links will appear without explanation and therefore less likely to be clicked.
The alternative – opting out of being indexed by Google and not appearing in the search engine at all – could be devastating to their business, they said.
“At least for now, that's not possible,” said Len Turiano, head of product at Gannett, the largest newspaper publisher in the United States.
But he said AI Overview “would do huge damage to everyone except Google, especially to consumers, small publishers and businesses large and small that use its search results.”
Google said its search engine continues to drive billions of visits to websites and provides value to publishers, and that it doesn't show AI summaries when it's clear users are looking for current events.
In an interview before the launch of AI Overview, Liz Reid, vice president of search at Google, said there had been signs of hope for publishers during the tests.
“We're finding that people continue to click on links in the AI Overview,” she says, “and websites that appear in the AI Overview actually get more traffic than websites that are just the traditional blue links.”
Reid wrote in a blog post Thursday afternoon that Google would be limiting search results to a narrower scope after some significant errors in AI Overview, but added that the company is still working to improve the system.
AI-generated summaries are the latest area of tension between tech companies and publishers, and the news site's use of articles has also sparked legal battles over whether companies like OpenAI and Google violated copyright law by taking content without permission to build AI models.
The New York Times sued OpenAI and its partner Microsoft in December, alleging copyright infringement of news content for training AI systems and services. Seven newspapers owned by Media News Group and Tribune Publishing, including the Chicago Tribune, have filed similar lawsuits against the same tech companies. OpenAI and Microsoft have denied any wrongdoing.
AI Overviews is Google's latest attempt to catch up with rivals Microsoft and ChatGPT developer OpenAI in the AI race.
More than a year ago, Microsoft built generative AI into the core of its search engine, Bing. Google initially took a more cautious approach, fearful of dipping its toes into its cash cow. But at its annual developers conference in mid-May, the company announced an aggressive rollout of its AI capabilities, saying the technology will be available to more than a billion people by the end of the year.
AI summaries combine statements generated from AI models with snippets of content from live links on the web. Summaries often include excerpts from multiple websites, citing sources, giving users a comprehensive answer without having to click through to another page.
Since its launch, the tool has not always been able to distinguish between accurate articles and satirical posts, sparking an online uproar when it recommended users put glue on pizza or eat rocks to get a balanced diet.
News organizations said in interviews that it's too early to tell if there's been any change in traffic from Google since the introduction of AI Overview, but the News/Media Alliance, a trade group representing 2,000 newspapers, has written to the Department of Justice and the Federal Trade Commission urging them to investigate Google's “misappropriation” of news content and block the company from rolling out AI Overview.
Many publishers say the move underscores the need to build more direct relationships with readers, whether that be by encouraging more to sign up for digital subscriptions or visiting their sites or apps directly, and reducing reliance on search engines.
The Atlantic's chief executive officer, Nicholas Thompson, said the magazine is increasing investments in all areas that directly connect with readers, including email newsletters.
Newspapers including The Washington Post and Texas Tribune are turning to Subtext, a marketing startup that helps businesses connect with subscribers and viewers through text messaging.
Subtext Chief Executive Mike Donahue said media companies are no longer chasing the biggest viewership but instead striving to keep the biggest fans engaged. His client, the New York Post, allows readers to exchange text messages with sportswriters as an exclusive perk.
Then there's the copyright battle. In an unexpected turn of events, OpenAI, which developed ChatGPT by scraping news sites, has begun striking deals with publishers. It has said it will pay companies like News Corp., which owns The Associated Press, The Atlantic and The Wall Street Journal, for access to their content. But Google, which helps publishers monetize with its advertising technology, has yet to strike a similar deal. The internet giant has long resisted calls to pay media companies for content, arguing that such payments would undermine the very essence of the open web.
“There's no running from the future. It's the future,” said Roger Lynch, chief executive of Condé Nast, publisher of magazines including The New Yorker and Vogue. “I'm not disputing whether the future will or should happen, but that it should happen on terms that protect creators.”
He said search remains “the lifeblood and the majority of traffic” for publishers, and suggested a solution to their woes could come from Congress: He's urging lawmakers in Washington to clarify that using content to train AI isn't “fair use” under current copyright law and requires a license fee.
The Atlantic's Thompson, who announced the OpenAI deal on Wednesday, still wants Google to compensate publishers as well. In the meantime, he said before the AI brief was released that The Atlantic wants to join Google's brief “to the extent possible,” despite the industry's concerns.
“We know that as Google makes this transition, we will see a drop in traffic,” he said, “but we believe that by participating in the new product, we will be able to minimize that drop in traffic.”
David McCabe Contributed report.
