Google AI summaries deepen pressure on search-dependent publishers

By Olivia Hart

Yahoo CEO Jim Lanzone cited traffic drops of 60% to 70% as AI answers reduced reasons to click, although he named no publishers and supplied no measurement period. At the Paley International Council Summit in New York, Lanzone called the problem “an existential issue for publishers,” TheWrap reported. The figure is an executive warning, not an independently verified industry benchmark. The reporting comes from The Verge’s Sunday Stepback newsletter by Emma Roth, TheWrap’s summit coverage and court documents unsealed last month.

What changed in Google Search?

Google launched AI Overviews in 2024. The feature places an AI-written summary of web pages at the top of results, with links to the sources the AI system judges most useful below it. Readers searching for news now see the summary before the links.

For years, publishers optimized their pages for Google. Google received content for its search index; publishers received exposure and clicks. The Verge also reported that Google changed how it ranks results. The report gave no rules, affected queries or before-and-after data, so the ranking effect remains unquantified.

The mechanism is straightforward: Google retains access to publisher content, while its answer layer can satisfy many readers before the publisher’s link receives a visit.

The evidence carries different weight. The 2024 launch and Google’s source-link announcement are product facts. The ranking account is reported but unquantified. Lanzone’s traffic range is a warning without named publishers or dates.

Takeaway: Google still gets indexed content, but an answer can intercept the visit that once followed a source link.

How much traffic is at risk?

Lanzone described the first 27 or 28 years of the internet as a bargain: publishers allowed search engines to crawl their sites, and search engines sent visitors. AI answers, he said, give people fewer reasons to click through.

His figure needs a strict label: an attributed traffic claim without a publisher list, analytics evidence, baseline or measurement period. Put simply, 60% to 70% would equal six or seven of every ten visits, but it is not a measured benchmark.

For online publishers and news sites that depend on Google referrals, the mechanism is direct: an AI answer can satisfy the query before a publisher link receives a visit. The source does not quantify effects on other site categories.

Lanzone said no business could operate that way. He also called on AI companies to do more to link readers back to their sources.

Takeaway: The warning shows severe exposure for search-dependent publishers, not a verified market-wide decline.

Is Google sending readers back to publishers?

TheWrap reported that Google announced changes last week to make links to original sources more prominent in its AI search features. That is an announced Google product change, but the source provides no rollout scope, implementation details or click-through results.

Google also runs a News AI licensing program with more than 400 participating publications, including CNN, Axios and the Financial Times. That establishes program scale, not referral gains.

Together, the two Google moves address publisher exposure, but neither establishes whether AI referrals can replace the organic clicks lost since summaries moved above links.

Takeaway: Source links and licensing can widen the response, though neither proves that traffic has recovered.

What do the court documents show?

Court documents unsealed last month contained an internal warning from Nick Turley, OpenAI’s head of ChatGPT. Turley warned that publishers faced an existential threat as AI products replaced visits to their sites.

The document records a concern inside OpenAI, yet it does not quantify publisher losses, identify a particular product action or establish the size of any traffic decline. Even so, it differs from an unattributed market estimate because it names the speaker and places the warning in court records.

Takeaway: The papers confirm the strategic concern, not the size of the traffic loss.

Can AI search return traffic instead of absorbing it?

Yahoo launched Yahoo Scout in January. The AI search tool gives answers while linking to the sites that supplied the information. Lanzone said Yahoo wants to return to its roots as a gateway to content across the internet.

That design puts source links inside the answer experience rather than treating them only as external destinations, but its referral effect remains untested without audience or publisher data. The source provides no user count, referral data or publisher results.

Takeaway: Yahoo Scout offers a referral-first model, but its audience impact is not yet measurable.

What does the open-web argument want?

Rose Wang, chief operating officer of Bluesky, described the current web as “technofeudalism.” Large platforms, she said, act like landowners who collect payments from creators working on their property.

Wang said users should have more control over their identities and data. She also described Meta CEO Mark Zuckerberg, OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei as monarchists with differing views of the internet’s future.

Her prescription was direct: companies that support the open web should send traffic to one another rather than depend on the largest platforms that concentrate distribution. That approach seeks to redistribute referrals through cooperation. It does not answer how AI search interfaces will change click behavior.

Takeaway: The open-web debate offers a distribution strategy, not evidence that publisher traffic has stabilized.

What should publishers monitor next?

Publishers should separate claims from operating data:

• Google referrals: Whether click-through changes follow the reported ranking update and more prominent source links.
• Licensing: Whether participation expands beyond more than 400 publications and participating outlets report referral gains.
• Yahoo Scout: Whether citations to supplying sites produce measurable traffic rather than attribution alone.
• AI summaries: Whether Google keeps summaries above links and how much attention the source links receive.

The next test is measurable: whether Google’s more prominent source links and Yahoo Scout’s citations restore referral traffic at meaningful scale across participating publishers.