The UK Competition and Markets Authority just did something no regulator has done before. They separated your right to display content from the right to use it for AI training. Google must give every publisher two independent controls by December 3rd, 2026.
This is not about fairness. It is about ownership.
When I ran due diligence at Hartford Steam Boiler for Munich Re, one principle governed every deal review: who owns the asset? Not who uses it. Not who built it. Who owns it. That question is now the most important question in digital marketing.
For three years, publishers lost that question entirely. Google scraped their content, showed summaries in search results through AI Overviews, and fed the material into model training and fine-tuning. Publishers got zero control. Zero visibility into what Google extracted. Zero negotiating power. The situation was asymmetrical in Google's favor by design.
The CMA fixed that. On June 3rd, 2026, they issued a binding conduct requirement under the Digital Markets, Competition and Consumers Act. It covers AI Overviews, AI Mode, Gemini, and Vertex AI. It applies at domain level and page level. It includes an anti-retaliation clause: Google cannot penalize your traditional search rankings if you opt out.
That last part matters. For years, publishers faced a false choice: accept AI scraping or vanish from search. The CMA eliminated the false choice.
This ruling separates content display from training rights in a way that forces real structural change. Publishers can now control whether their content feeds Google's AI systems independently of whether Google's search algorithm ranks them. That independence is fundamental. It means opting out of AI training does not trigger algorithmic retaliation.
Here is what you must do immediately.
First: Take inventory of your AI exposure right now. Google's testing began immediately on UK-based media sites. Main controls roll out December 3rd. Page-level grounding controls follow March 3rd, 2027. You have nine months to audit which content creates real value and which content you want to protect. Start this work today. Do not wait for December.
The CMA's timeline is not negotiable. December 3rd means December 3rd. Every week before then is time to audit your content portfolio. Which pieces have you written that ended up in Google's training data? You will not know for certain. But you can estimate. Think about your evergreen content. Think about your technical pieces. Think about your reporting that other outlets cited. That is the content Google trained on.
Second: Prepare your opt-out strategy. The controls give you two separate controls. One controls whether your content appears in AI-generated summaries at the top of search results. The other controls whether Google uses your content to fine-tune AI models. These are separate decisions. You can allow summarization while blocking training. You can block both. You decide the configuration that serves your business.
Consider this carefully: if your content has high search value but low training value, you might allow summaries while blocking fine-tuning. If your content is specialized and proprietary, you might block both. If you are negotiating a licensing deal, you might block everything until Google agrees to terms. The point is your decision, not Google's default.
Think about your site structure. A news publication might block AI training on breaking news but allow it on evergreen guides. A financial services publisher might protect analysis and research but allow summaries of commodity information. A software documentation site might require payment for training access but allow summaries in search. Your strategy should vary by content type.
Third: Demand attribution and links everywhere. The CMA requires Google to include "clear links and proper attribution" in AI-generated results. This is not optional. If Google summarizes your reporting, the user sees your URL and your byline. Check Search Console regularly. Demand compliance verification. The CMA will require Google to submit compliance reports every six months in year one. Those reports are public. Use them to verify Google is following the rules.
Attribution matters for two reasons. First, it protects your brand and drives traffic. Second, it proves what content Google used. Those impressions become your negotiating data. If you can see exactly which pieces Google summarized 100,000 times, you know those pieces have training value.
Fourth: Use opt-out controls as negotiating power for licensing deals immediately. Publishers from The New York Times to Axel Springer have negotiated AI licensing deals with Google. Before June 3rd, those negotiations happened from weakness: Google could scrape with impunity. Now publishers hold the switch. Tim Cowen of the Movement for an Open Web warned that verification might not happen until late 2027. Do not wait for perfect verification. If you have meaningful traffic, Google wants permission to use your content. Make them pay for it now.
The licensing conversation changed fundamentally on June 3rd. You went from a supplicant begging for scraps to a property owner Google must negotiate with. Price accordingly. If your content generates millions of impressions, model fine-tuning built on that content has real value. Own that value. Monetize it.
The CMA ruling shifts the entire power dynamic of content licensing. Before this ruling, publishers had no choice. Google could train on publisher content without permission because the relevant laws did not clearly establish content ownership boundaries for AI training. Publishers could complain. Google ignored them. No remedy existed.
Now a remedy exists. The remedy is your opt-out right. Use it as currency. Publishers with significant traffic (think sites getting 5 million monthly impressions or more) should calculate their training value. How much does a model improve if trained on your content? Nobody knows. But Google's own behavior reveals the answer: if Google will negotiate payment, the content is valuable.
Start with a floor price. Your content. Your rule. If Google balks, opt out. See what happens to their model performance. That real-world data becomes your pricing justification.
Many publishers will discover that their niche content has outsized training value. A financial analysis site, a medical publishing operation, a technical documentation archive—these assets train models far more effectively than generic content because they are so specific and dense with expertise. Use that specificity in pricing. You are not selling inventory. You are selling training data with professional provenance.
Fifth: Track your metrics obsessively. The CMA requires Google to publish disaggregated engagement data in Search Console: impressions, click-throughs, CTR. This baseline data proves your current value to Google. Get the numbers now. Create a baseline report. Use it in licensing negotiations. Use it to understand the real cost of saying no to any request from Google.
You should know: if Google trains models on 10 million impressions of your content monthly, that training has a floor price. That price is negotiable. You need data to set that floor.
The Search Console data will show you three things. First, how much traffic your content attracts. Second, which pieces Google values most. Third, which pieces will be most damaging to Google's models if you block them. These three data points become your pricing anchor. Do not price based on hope. Price based on data.
Sixth: Watch the global rollout with extreme attention. Google said it will test in the UK first and then roll out globally. The CMA ruling covers Google's UK operations. But Google's parent company operates in 190 countries. Watch for: Does Google implement these controls in the EU? In Canada? In Australia? Will the CRTC, the ICO, or the Bundeskartellamt demand the same rights? This ruling will echo across jurisdictions.
Each jurisdiction that adopts similar rules increases your negotiating power with Google globally. If 50 percent of Google's traffic comes from jurisdictions with opt-out controls, Google's ability to train models without permission drops significantly.
Ownership matters because it creates power in negotiation. When you own the asset, you negotiate from strength. When someone else owns it, you negotiate from weakness or not at all. This is not theoretical.
Consider the model licensing deal structure: you own content. Google wants to use content in training. Google offers you $10,000 per month for unlimited training access. Without the CMA ruling, you had no option but to take it or get zero. With the CMA ruling, you can say no. That "no" is worth money to Google. The difference between your take-it-or-leave-it offer and your negotiated deal is pure negotiating power.
For three years, Google owned the question of how your content would be used in AI training. You had no say. The CMA took that power back and put it in your hands on June 3rd, 2026.
Do not waste it. Audit your content today. Understand your AI exposure. Decide what you want to protect. Opt out where it makes sense. Demand attribution everywhere. Negotiate licensing for remaining uses. Set pricing based on real value, not fear.
The principle that governed every deal review at Hartford Steam Boiler still holds: ownership beats everything else. The CMA finally made that principle operational in AI. Your job is to act on it before December arrives and you are asked why you did not.
*Jeff Barnes, MBA has no personal position in any company, tool, or platform named in this article. demg.ai provides marketing education and systems for owner-operators, not investment advice. Past performance does not guarantee future results.*