“What difference will AI summaries make to news media?” I asked Google, choosing the search engine’s new AI mode.

In a few seconds I received a summary paragraph, bullet-point arguments in four boxes titled “Benefit, Risk, Challenge and Opportunity”, and a summary of “the strategic outlook for news media”. There was enough material for me to copy and paste an entire slide presentation, had I needed one.

It also gave me links to 18 of the sources it had used, and a footnote that “AI responses may include mistakes.” Yes: as one of the cited articles found in February 2025, over half of AI chatbot summaries had problems with accuracy. But that’s not the only problem with Google’s AI tools.

The Financial Times has reported that publishers are scrambling to counter the “Google Zero” threat after a sharp drop in web traffic. Since Google introduced AI Overviews in May 2024, as an addition to standard search, users who see it are less likely to click on a link to a source article — half as likely, in fact. Pew Research found that encountering an AI summary cut the click rate from 15% to 8%. Members of digital media trade body DCN (Digital Content Next) reported that, in the week following the introduction of Google AI Overviews, news brands suffered a 16% drop in site visits via Google searches.

This is bad news for the media organisations producing the news that Google’s AI is summarising. If readers never bother to click through, they’re neither paying for content nor seeing adverts. Without that revenue, journalists can’t be paid, and the sources will run dry. No wonder, then, that the owner of the Daily Mail has recently called for a crackdown on the tech company.

Google is not the only platform using generative AI summaries, of course, but as a search and advertising giant it has strategic market status. The Professional Publishers Association (PPA) has submitted data on the fall in traffic to its members’ sites as evidence to the UK’s Competition and Markets Authority (CMA), demanding that something be done to staunch the bleeding.

This isn’t the first time media companies have accused digital platforms of a parasitic relationship to their original work. When Facebook introduced Instant Articles in 2015, promising to make publishers’ content easier to access on mobile devices in particular, many worried that it handed more power, and ultimately more editorial control, to social media platforms. But Facebook did at least share revenue with the originators of the material.

Jason Kint, CEO of DCN, says the difference now is that AI summaries effectively replace the content from which they draw, discouraging the reader from ever following through to read the original material. Like AI-generated music or visual artwork produced by training on human creations, the AI-generated news summary first devours human work and then takes its place. However, it’s hard to see AI-summarised news surviving for very long without a constant supply of fresh journalism to summarise.

What can media organisations do? Refusing permission to include their output in AI summaries would also remove their links from the results page altogether, resulting in an even bigger drop in traffic to their sites. The obvious answer is licensing agreements, giving Google AI and its ilk access to the original content in return for a share of revenue.

OpenAI already has licensing agreements with large media organisations, allowing it to train its Generative AI programmes on their output. Google is perfectly placed to make similar agreements, which would allow it to use publishers’ output as both training data and raw material for news summaries that keep readers on its own pages.

Will such agreements safeguard the independence and financial security of news organisations? Of course not. But they may not have much choice. As Google AI mode told me: “For the news media, AI summaries are part of a larger power struggle with tech platforms over content and audience.”

---

Timandra Harkness presents the BBC Radio 4 series, FutureProofing and How To Disagree. Her book, Technology is Not the Problem, is published by Harper Collins.

TimandraHarknes

---