In a different phase of testimony in the federal antitrust case that could redefine online searching, Eli Collins, Vice President, Google DeepMind, mentions that Google training its search-related AI products AI Overviews using web content might be permitted even if publishers have clearly opted out of having their data used in general AI training.
The Legal Loophole
When asked whether Google Search can continue using content from sites that have opted out of AI training via existing protocols, Collins said:
“Correct — for use in search.”
This distinction means that Google’s AI models embedded within Search — like its Gemini-powered summaries and featured answers — can still access and learn from web content, regardless of AI opt-out settings intended for Google’s broader AI lab, DeepMind.
What This Means for Publishers
For website owners and media companies, this may feel like a bait-and-switch. Publishers can only stop Search from using their content if they completely opt out of search indexing — a move that would eliminate their visibility on Google altogether.
According to a Google spokesperson, publishers must rely on the robots.txt protocol to manage access, but this system doesn’t distinguish between AI and search purposes.
In other words, if you’re on Google Search, your content can be used to train Google’s search AI, whether you like it or not.
Why This Is a Big Deal
Many publishers have raised alarms that Google’s AI-generated answers reduce click-through traffic — offering users full responses without visiting the original sources. Now, they’re learning that Google is using their own content to power those summaries, even without consent under AI-specific opt-out rules.
The practice has landed squarely in the crosshairs of the U.S. Department of Justice, which has already won a rulingthat Google illegally monopolized the search market. The DOJ is now proposing aggressive remedies, including:
- Forcing Google to sell Chrome
- Banning default search payments on phones and browsers
- Restricting AI-related advantages tied to its dominance
AI and Antitrust: The New Frontier
The implications of this case go well beyond web traffic. Regulators argue that Google’s dominance in search has given it an unfair head start in training AI products like Gemini, especially as those tools integrate deeper into services like Google Search and Android.
The DOJ’s goal is to restore competition not just in search, but in the emerging AI economy — a sector Google is already poised to dominate thanks to its access to massive amounts of user behavior and content data.
Judge Amit Mehta, who is overseeing the trial, will now decide what structural changes, if any, Google must make to curb its dominance and give rivals a fair shot in both search and AI.
