Publishers test AI search strategies as Google referrals fall 40%

Google Search referrals to hundreds of publishers fell 40% from July 2025 to July 2026. The decline was twice as steep as the previous year, according to Chartbeat’s annual publisher playbook.

Against that backdrop, a Reuters Institute report examined how publishers are approaching artificial intelligence search. Some are testing generative engine optimization, or GEO, while many remain focused on direct audience relationships.

The report found no settled playbook for improving visibility in AI-generated answers. Ezra Eeman, director of AI in Media at WAN-IFRA, said he has seen “countless playbooks” offering contradictory advice.

The sources cited most often by AI tools also keep changing, Eeman said. Wikipedia led at one point, followed by Reddit and YouTube. The patterns also differ across platforms.

Shelby Blackley, The Athletic’s director of global SEO and AI search, pointed to the limited available data. “We have such little data across the industry that it’s actually annoying,” she said.

Consultant Clara Soteras said the process is not completely opaque. Publishers can monitor prompts, brand mentions and content formats through testing, she said. Still, much of the work remains experimental.

Axios is testing likely user questions and tracking how often its journalism appears in AI responses. Ben Berkowitz, who oversees newsroom AI operations, said the company focuses heavily on authority.

Reuters Institute described that work as experimental because identical questions can produce different answers and citations. Berkowitz also said Axios’ Smart Brevity format “is incredibly LLM-friendly.”

Visibility does not necessarily translate into traffic.

A Pew Research Center study examined more than 68,000 Google searches from 900 U.S. users. Just 1% ended with a click on a source cited in a Google AI Overview.

Eeman argued that citations could still build recognition and strengthen a publisher’s position in commercial discussions. He offered one hypothetical pitch: “We are in 80% of answers. We’re always there. You need us.”

Publishers also face decisions about whether AI systems can access their content.

Reuters Institute research found 48% of major news sites across 10 countries blocked OpenAI crawlers by late 2023. Blocking retrieval crawlers can also reduce a publisher’s chance of appearing in relevant AI answers.

Eeman suggested publishers could take a selective approach. They could protect premium journalism while allowing access to freely available material.

The Guardian’s Jessie Willms cautioned against treating GEO as a replacement for traditional search engine optimization.

“I don’t think it’s necessarily priority number one if you haven’t already done all the basics,” she said. Willms described GEO as another layer on top of established SEO practices.

Soteras also questioned whether traditional audience metrics capture AI visibility effectively. Many newsrooms still emphasize unique users and page views, she said.

That challenge comes as publishers respond to broader search traffic declines. Many are also investing in direct audience relationships through newsletters, events and other products.

For smaller outlets, Eeman sees one possible advantage. They can surface when they provide information few other sources can answer, particularly on local topics.

The broader takeaway is less dramatic than a wholesale newsroom pivot. Some publishers are experimenting with AI visibility while still prioritizing SEO and direct relationships.

For now, GEO remains an additional layer — not a replacement for the audience strategies publishers already use.