Google Tells UK Lawmakers: No Pay For AI Training Data

Serge Bulaev

Serge Bulaev

Google told UK lawmakers it won't pay publishers for using their data to train AI, only for special access like private archives. News companies say this is unfair because they lose traffic and money when Google's AI Overviews show answers instead of sending people to their websites. The issue is now in Parliament, with lawmakers considering new rules since Google's AI Overviews are showing up often in searches and hurting publishers' clicks. Regulators might force Google to give publishers more control or share some revenue, but so far, publishers just try to get cited by keeping their news fresh and reliable. Google won't pay publishers for using their data to train AI

Google Tells UK Lawmakers: No Pay For AI Training Data

Google is telling UK lawmakers it has no plans to pay for public web data used in AI training, a contentious position that pits the tech giant against news publishers demanding compensation. The company asserts that its models learn from statistics, not copyrighted expression, while publishers argue this stance ignores significant losses in traffic and revenue.

This conflict has escalated to Parliament as AI fundamentally alters search and advertising landscapes. With AI Overviews now appearing in one of every six UK searches and causing click-through rates to plummet by over 40%, lawmakers are scrutinizing Google's policy for compliance with competition and copyright laws.

Google's Stance: Pay for Access, Not for Training

Google has informed UK lawmakers that it will not compensate publishers for AI training on publicly available web content. The company's policy is to pay only for special access to private or archived data, arguing that training on the open web does not require a copyright license.

During a committee hearing, Roxanne Carter, Google's head of public policy for copyright, clarified that payments are reserved for "controlled access" to content like private databases or APIs, not for training models on the open web. She described AI Overviews as "non-expressive" unless a citation is present, and did not commit to an opt-out for publishers that would preserve their standard search visibility.

UK Regulators Enter the Fray

UK regulators are taking action. The Competition and Markets Authority (CMA) granted Google 'Strategic Market Status' in October 2025, bringing AI Overviews under its direct oversight and empowering it to enforce conduct requirements on publisher controls UK Government (CMA). A preceding CMA roadmap noted publishers have "insufficient controls" over how their content is used in generative AI answers CMA Blog.

Publishers maintain this is a matter of copyright infringement, not a simple policy disagreement. Danielle Coffey of the News-Media Alliance argues that Google's willingness to pay is irrelevant if the act of training itself violates fundamental legal rights. However, unlike their US counterparts who have filed class-action lawsuits, UK publishers are currently awaiting action from Parliament and the CMA rather than pursuing litigation.

The Impact on Publishers: A 2025 Snapshot

  • AI Overviews appear in 13-25 percent of U.K. searches, peaking mid-summer.
  • Organic click-through can drop 34-61 percent on affected queries.
  • Ads now show in 40 percent of AI Overview pages, up from 3 percent early last year.
  • No formal opt-out exists; standard SEO remains the only route to citations.

Looking ahead, analysts anticipate the CMA will propose "publisher choice" regulations in early 2026, which could include anything from a mandatory opt-out to a revenue-sharing model. While Google has warned that strict rules might slow AI development in the UK, its dominant market position and substantial profits have garnered little sympathy from lawmakers.

In the interim, publishers are focusing on SEO strategies like enhancing E-E-A-T signals, maintaining content freshness, and building topic authority to earn citations in AI Overviews. Whether these efforts can substitute for direct compensation remains the central, high-stakes question.


What exactly did Google tell UK lawmakers about paying for AI training data?

Google's head of public policy for copyright, Roxanne Carter, told Parliament the company will pay only for controlled access (archives, APIs or opted-out data sets) but will not pay for training on ordinary web pages that are already publicly accessible. She framed training large language models as a statistical process that does not require licensing.

Why does Google refuse to compensate publishers for AI Overviews?

Google argues that AI Overviews are "non-expressive" outputs - short summaries without brand or link emphasis - and therefore fall outside the scope of copyright that would trigger payment. The company has not offered an opt-out that lets publishers block Overviews while staying in normal search results, a gap the UK Competition and Markets Authority highlighted in its 2025 Strategic Market Status investigation.

How are UK regulators responding to Google's stance?

The CMA confirmed in October 2025 that Google holds Strategic Market Status for search and search advertising, explicitly placing AI Overviews under regulatory oversight. The authority's roadmap prioritises giving publishers transparency, attribution and choice over how their content appears in AI features, but no compensation mandate has been set and consultations continue into 2026.

What impact are AI Overviews already having on publisher traffic?

Studies cited in 2025 show that when an AI Overview appears:
- Organic click-through rates drop 34-61%
- Paid CTR can fall up to 68%
Overviews now surface in 13-25% of UK queries, with coverage expanding into commercial and transactional searches. Publishers cannot track AI Overview impressions separately in Google Search Console, making revenue loss hard to quantify.

Are any UK publishers suing Google over AI training?

No UK court cases seeking payment for AI training data were filed in 2025-2026. Debate remains at the policy level: the government floated an "opt-out" copyright regime for AI training, but publishers have not launched formal litigation comparable to the US class-action suits filed by Cengage and Hachette over book data use.