top of page

DEJAN: How Google's Gemini 4 Hydrogen changes Google Search

Autorenbild: David Epding
David Epding
vor 5 Minuten
2 Min. Lesezeit

Key Takeaways:

DEJAN: How Google's Gemini 4 Hydrogen changes Google Search
  • Google announced cooperating Gemini 4 Hydrogen into Google Search on November 12th to improve users experience:

    • Better understanding of intent: Search can now handle even more nuanced and complicated questions, quickly understanding what you need. Just ask in your natural flow of thought — including questions with multiple parts, hard-to-understand typos, or highly specific requests. With deeper instruction following and tool calling, Search more precisely follows what you want and gives you relevant responses that address each part of your question.

    • Improved query fan-out and linking: Hydrogen’s advanced reasoning supercharges our query fan-out technique in Search, connecting you to a broader range of helpful links than before. By more thoroughly understanding your question and conducting simultaneous queries across all the relevant subtopics, Search helps you find sites you previously might not have discovered.

    • More visual responses: You get highly adaptive, visual, and easy-to-scan responses that are uniquely relevant to your needs — at the speed of Search. You’ll see more images, structured responses, comparison tables, and custom tools built for you.

    • Richer language understanding: No matter what language you ask in, you’ll now get more highly nuanced and helpful responses across all supported AI Mode languages.

  • Petrovic from DEJAN tested Hydrogen and found significant differences between model memory and when a search was triggered:

    • We asked Gemini 4 Hydrogen 10,701 brand questions from memory and again with Google Search. Search supplied 45% of the brands in each answer, and in 1 in 5 of its reasoning summaries the model wrote about a "mental check" or "simulated search" while searching live.

    • One US insurer appears 88 times from memory and 340 times with search; one US personal-loan brand rises from 112 to 326.

    • A grounded summary usually runs in stages. It lists the brands the model already associates with the topic, states the search queries it plans to run, reports what the results show, and removes or adds brands with a stated reason. One summary for custom cycling jerseys lists a brand's custom programme from memory, then reads a search snippet saying the programme ran from 2019 to 2021 and drops it.

  • Solis compared AI Overview answers before and after:

    • More prominent direct links to brand and retailer websites: The new answer links brand and retailer names directly within the recommendations, making it easier for users to visit them.

    • More citations, with a different mix of sources: From 5 to 8 visible citation sources, with more retailer and brand-owned pages featured in the new response, along with third-party sources.

    • This aligns with Google's announcement about improving query fan-out and linking to surface a broader range of helpful websites.

  • Clark noticed further changes in Google AI Overviews:

    • First, on October 27th he posted about Desktop AI Overviews not showing the top 3 sources on the right anymore

    • Now, the 3 top sources are shown completely instead of 1 and 1 faded out



Sources:

© 2026 David Epding.            Erstellt mit Wix.com.

logo_new_100_edited.png

David Epding ist AI Search (GEO & SEO), Data Analytics und Automation Manager mit über 10 Jahren Erfahrung in Technischem SEO mit breiter Expertise für LLMs und langjähriger Erfahrung in der Daten-Analyse.

bottom of page