Since AI Overviews launched, search and publishing professionals have been paying close attention to how AI companies should handle the content used to train their models. Google has now shared its stance. It emphasizes fair use and provides options for opting out, while also highlighting paid agreements for specific situations.

In a policy paper published June 25, Google shares that training models on publicly available web data is considered a “transformative, non-expressive use” that should remain protected under fair use in the U.S. The company highlights opt-out controls and existing copyright law as their main solutions for addressing publisher concerns.

The paper, “A Pragmatic Approach to AI Governance in America,” gathers together the points Google has shared previously. It comes at a time when regulators and publishers are pushing for more, seeking not just opt-outs but also clearer attribution and sometimes even compensation. For publishers figuring out how to manage AI access to their content, it offers helpful insight into where Google stands.

Google’s Copyright Position

Google likens AI training to “an art student taking inspiration from walking through a gallery.” It also suggests that the same level of protection should be extended internationally through text-and-data-mining exceptions.

For site owners who don’t want their content used, Google recommends using machine-readable controls like Google-Extended in their robots.txt. When AI outputs copy existing work, the solution isn’t about filtering to judge if an output is “too similar,” but relies on well-known notice-and-takedown processes, as outlined in the paper.

Google is also looking into new ways to create value, such as partnering with websites that provide content helping to keep AI responses up-to-date and accurate, and deals to pay for access to specialized, non-public content. The paper doesn’t specify any particular programs, terms, or timelines.

Where The Position Lands

This month, the UK’s CMA introduced a new conduct requirement that gives websites the option to opt out of AI search features and requires Google to attribute publisher content. The regulator mentioned that this measure is intended to help boost publishers’ bargaining power. Google has already started testing an opt-out toggle, though the reports available to publishers to help them decide haven’t yet included click data.

US publishers are making their stance even clearer. Digital Content Next recently sent a cease and desist letter to the Common Crawl Foundation, emphasizing that “copyright law is not an opt-out regime.” This means that scrapers should seek permission before using content, rather than publishers having to request to be excluded. This perspective directly challenges the opt-out model discussed in Google’s paper.

Why This Matters

The paper highlights Google’s stance as policymakers consider new rules. Google is advocating for keeping its current approach unchanged.

Publishers and regulators are seeking more than what the paper currently provides. They’re requesting compensation, permission-first scraping, and detailed click-level data. In response, the paper offers controls and handles negotiations on an individual basis.

Looking Ahead

These are policy positions, not product commitments. The grounding partnerships and content deals Google mentions could influence how value reaches publishers, but the paper leaves the details flexible. Keep an eye on whether Google links programs, terms, or figures to the value-exchange language it’s currently including in its policy documents.

Featured Image: FotoField/Shutterstock