Google makes Gemini 4 Hydrogen the default model in Search's AI Mode worldwide

What Google actually rolled out

On Thursday, November 12, 2026, Google began serving Gemini 4 Hydrogen in Search. As PPC Land reported, it is now the default model in AI Mode, Google's conversational search tab. For harder questions, it also powers AI Overviews, the summaries shown at the top of ordinary results.

Google's own framing is blunt. Nick Fox, who wrote about the launch, said, "It's the biggest model improvement to Search yet." He added that it is available to all users, globally.

That is a broad claim, and the reports we have do not say what it means in practice for every country, language or account type. Treat the rollout as something to check in your own account rather than assume.

What Hydrogen is supposed to do better

Google's launch post names four areas of improvement:

• Making sense of complicated or misspelled questions.
• Splitting a question into subtopics and returning a wider range of links.
• Answering with tables and custom tools, not only paragraphs of text.
• Giving more nuanced replies in every language AI Mode supports.

Other coverage describes Hydrogen as a model aimed at efficient reasoning on everyday tasks rather than at topping a leaderboard. That fits a product meant to answer routine questions quickly, though none of our sources publish benchmark scores, usage limits or pricing for it.

Benchmarks deserve some skepticism in any case. A separate Google DeepMind and Artificial Analysis report, covered by PPC Land, ranked Anthropic's Claude Fable 5.2 first and Google's Gemini 4 Carbon second. The scores were 73.26% against 71.89%, a gap the report calls statistically insignificant. Fable's average cost per task was $35.28, against $5.63 for Carbon.

The Nvidia claim we are not repeating as fact

PPC Land's headline pairs the Hydrogen launch with a claim that Jensen Huang has quit as Nvidia's chief executive. The article describes a resignation and an interim replacement.

We have no Nvidia statement or second outlet in our set that backs this up, so we are treating that part as unverified. The Hydrogen rollout and the Nvidia story are two separate claims, and only the Search launch is supported by other reports. If you budget for AI hardware or tools that depend on Nvidia chips, wait for the company itself or several independent outlets before changing any plans.

Why distribution matters for the open web

Putting a stronger model in front of every searcher means more questions get answered on Google's own page. The traffic data suggests this is already biting. SparkToro, using clickstream data from Similarweb, found that visits to the top one million websites have fallen 12.2% since early 2022. Quarterly visits slid to 1.970 trillion, down from 2.243 trillion, even as the online population grew about 24%.

News publishers feel it most. Chartbeat figures cited by PPC Land put search's share of North American publisher pageviews at 12% in the third quarter, compared with 21% a year earlier.

Google is also nudging people toward Gemini elsewhere. It widened a beta that lets Gemini work with food, ride and reservation apps to five countries. It also published a Chrome guide pointing shoppers to Gemini for summaries of product reviews. PPC Land cited a UC San Diego study in which people who read AI review summaries said they would buy 84% of the time, versus 52% in a comparison group.

How to judge it for yourself

Google's launch claims are just that: claims. The useful test is whether Hydrogen does better on your own questions. Pick a few searches you have run before, especially ones with typos or several parts, and see whether the answers and links improve.

One caution for site owners: Google's new help page on identifying Search traffic says its method is not exact. It does not mention AI Mode or Gemini referrals, so measuring what Hydrogen sends you will stay murky for now.