Google's AI Answers in Search Just Lost Their Legal Shield

The Text Box That Changed Everything

You type a question into Google. Used to be, you'd get a list of blue links. Now, a written text box appears right at the top and hands you the answer directly. It's called AI Overview.

Useful, most of the time. Useful enough that almost nobody scrolls down to the sources listed below anymore. It's become the default reflex for hundreds of millions of people, and barely anyone stops to think of it as anything more than a shortcut.

That text box looks like a minor interface tweak. But in one week in July 2026, two European regulators decided it actually changes what Google is, legally speaking. On the 14th, Germany's media authority. On the 16th, the European Commission. Neither coordinated with the other, and both zeroed in on the exact same thing: neutrality.

Twenty Years of Neutrality as a Shield

To get why this matters, you have to go back to Google's old unspoken deal. For two decades, its defense boiled down to one idea: Google doesn't create content, it just sorts and links to it. There's a name for that status in law: a host. A host carries other people's information, the way the postal service carries mail. Nobody sues the post office over what's written in a letter.

That shield even has a technical nickname: "safe harbor." As long as Google was just pointing to pages, it stayed protected.

The trade-off was simple: stay neutral. Don't write your own content. Don't favor your own services. The pipe doesn't talk and doesn't pick sides, it just lets things flow through. That's exactly what AI just broke, from both ends at once.

Angle One: Google Becomes the Author

When the engine generates the answer itself instead of pointing you to a source, it's no longer carrying anything. It's writing. That's the finding from ZAK, Germany's media regulator commission, on July 14, in what are the first rulings of this kind in Europe. The target was AI Overview, but Perplexity's chatbot got hit with the same logic. The wording is blunt: AI-generated answers are "the provider's own content."

That's a heavy consequence. If it's your content, you're liable for it. The host privilege that used to shield Google no longer applies. Its answers now fall under media rules: accuracy, transparency about how links get picked, no discriminating against news sources.

ZAK's specific complaint says more than any lengthy statement could. Inside AI Overview, the AI-written text is displayed so prominently that the regular list of links gets pushed down the page. The result: news articles become harder to find, and publishers are already paying the price for it. To a media regulator, that's not curation anymore, that's an editorial call.

Angle Two: Google Becomes the Referee

Two days later, the European Commission hit the same wall from the other side. Its focus isn't content, it's power. Google controls access to the market: it decides who gets to reach users. In Digital Markets Act (DMA) terms, that makes it a "gatekeeper."

A gatekeeper that pushes its own Gemini assistant to the front and keeps the data to itself isn't neutral either. It's playing favorites. So on July 16, the Commission adopted two binding decisions. The first: Google will have to share its search data, queries, clicks, rankings, with rival search engines, and explicitly with AI chatbots doing search. Rollout starts in January 2027.

The second: on Android, rival assistants will have to get the same access Gemini currently enjoys on the phone. Eleven system functions are on the list, from voice activation to in-app actions. Deadline: the next major Android release, no later than August 1, 2027. In practice, you'll soon be able to launch ChatGPT or another assistant the same way you fire up "Hey Google" today.

Two Rulings, One Idea

Looked at separately, these two cases seem unrelated. One is about media law, the other about competition. But they're saying the same thing through two different doors. When Google generates the answer, it becomes the author, and answerable for what it writes. When Google ranks and promotes its own tools, it becomes the referee, and obligated to play it straight.

Author on one side, referee on the other. Either way, the "we're just a pipe" posture doesn't hold up anymore. And that shift matters, because neutrality was never just a talking point. It was a legal status, with real protections attached.

A middleman that just lets things pass through and a player that starts talking don't get judged by the same rules. Lose the first role, and Google risks losing the protections that came with it.

What This Doesn't Say Yet

Worth keeping straight: a direction isn't a verdict. Neither ruling is a fine. In Germany, the decisions aren't final and Google has said it's appealing. On the EU side, these are compliance measures, not financial penalties. The DMA's famous cap, up to 10% of global revenue, is real, but it's a threat for non-compliance, not a punishment that's already landed.

Google, for its part, isn't just stalling for time. It's making a real argument: sharing Europeans' search data and throwing the system wide open could put their privacy and security at risk. That's a legitimate debate, and it'll get settled in court.

Still, the line has moved. For twenty years, Google could present itself as a neutral pass-through for the web. AI turned it into something more powerful, and more exposed: a voice that answers and a hand that sorts. And in Europe, whoever talks and whoever referees is starting to have to answer for it.