Google’s Use of Publisher Data Faces Scrutiny in Antitrust Trial

In internal documents revealed during an antitrust trial, an executive from Alphabet Inc. acknowledged that Google could approach web publishers for their consent regarding the use of their content to enhance AI-generated search results. Alternatively, publishers could opt-out. However, the firm concluded that offering these options would complicate the AI training process tied to search functionalities.

According to the documents, which surfaced during ongoing legal proceedings, Google has established a “hard red line” where it intends to use publisher data to boost AI features, without providing any choices to the content creators. Instead of being transparent, the company plans to “silently update” its policies regarding publisher data, relying on a vague public announcement that won’t detail the use of this information, as noted by Chetna Bindra, a product management executive at Google Search.

Last year, a federal court ruled that Google operates an illegal monopoly, significantly strengthening its position in the AI landscape. Google maintains that it can utilize content meant for search engine results to inform the development of other AI-related products. The options available for publishers to shield their data from AI are essentially limited; they can only opt-out by withdrawing from search altogether.

Given that Google commands over 90% of the search market, skipping involvement in Google’s listings is not a feasible choice for many site owners. Many online creators depend on this traffic. The AI-generated overviews that answer queries directly can potentially reduce site visits, eliminating opportunities for ad revenue and product sales that come from user engagement.

Paul Bannister, the chief strategy officer at Raptive, highlighted that internal documents indicate Google recognized early on the possibilities for enhancing publisher control. “It’s a bit concerning,” he remarked, “as it shows they were aware of various options but opted for the most restrictive approach.”

In a recent trial in Washington, Judge Amit Meta is considering measures to restore competition in online searches. He is reviewing a set of recommendations proposed by antitrust authorities aimed at diminishing Google’s market advantages. Testimonies wrapped up on May 9, with a ruling on these proposals expected in August.

The Department of Justice’s proposal includes mandates for Google to utilize content from online publishers to train its generative AI models on a case-by-case basis.

Internal slides disclosed various options, such as allowing publishers to opt-out solely from certain AI features without entirely disappearing from search results. This could enable publishers to manage what content appears in AI Overviews, though their data would still be used for training purposes. One discussed option had publishers able to completely exclude their data from foundational use, which serves to ground AI responses in real-world sources.

Ultimately, Google decided against providing new choices for publishers, instead directing them towards existing options like the “No Snippet” choice that, when selected, would lead to less visibility on search pages and consequently, fewer click-throughs.

A Google spokesperson, answering inquiries about the court documents, stated that publishers have long maintained control over how their content is utilized in search. They insisted that the materials presented reflect early considerations that do not necessarily indicate feasibility or final decisions.

Documents introduced in court further included details on how company representatives might communicate information more vaguely. A document from April 2024 indicated that specificity might be addressed in future communications, while Google significantly integrated AI into its search functionalities at a developer conference a month later.

Since the introduction of AI Overviews, traffic to some publisher websites has noticeably declined. The long-term implications for publishers hinge on the ability of models to produce enough content to sustain their operations.

“If Google’s model progresses to the extent that human-created content is diminished, it’s like they’re signing their own death warrant,” said publisher Hartley Moy.

Moy noted the significance of the RAG technique, used by publishers to explore new revenue avenues and utilize content efficiently for search generation. She stressed that Google’s decision to remove discussions around RAG from negotiations is particularly crucial, as it solidifies Google’s market extensity while diminishing leverage for publishers in negotiations.

During the investigation, Google’s lawyers testified that creating multiple opt-out options for varying products is no easy task. They argued that each feature with generative AI could require distinct models, collectively leading to excessive complexity and costs in hardware and responsiveness. “Every feature needs its own model, and that’s not just costly but impractical,” one testified.