KorvaJoin the waitlist
Tech

Google Search Answers Messy Questions With Many Smaller Searches

The fan-out technique splits one tangled prompt into subqueries to surface pages a single search would miss.

AI image
TechCrunch · Google's latest Search upgrade promises better answers to your most complicated questions | TechCrunch

Google says Search is getting better at handling tangled questions with several parts, very specific requests, and confusing typos. The change is being released to Search users around the world.

The update is powered by Google's recent Hydrogen addition to its Gemini 4 line, which is designed to understand what you meant in your natural train of thought. Google says it is better at grasping intent, following instructions, and navigating hard-to-understand typos.

Search uses a method called fan-out to split a single prompt into multiple subqueries, or smaller searches run on your behalf.

Each subquery chases a different part of the question. Because each one looks for something slightly different, the combined results can include relevant websites you might otherwise have missed.

Hydrogen improves that process because it is better at addressing each part precisely and calling the right tools, Google says. In Search, it is meant for fast work across multi-step tasks, where the system carries out several actions on its own.

The practical takeaway is to ask the whole messy question as you think it, rather than trimming it into keywords. The system is built to do the splitting for you.

Questions people ask

What exactly is fan-out?

Fan-out means one prompt becomes several searches. Google runs those subqueries, gathers the different sets of results, and uses them to build a fuller answer with links.

Does this replace the list of website links?

No. Google says Hydrogen is meant to offer more helpful links alongside visual responses in AI Mode, including images, structured answers, comparison tables, and custom tools.

Why doesn't Search just use Google's most powerful model?

Google positions its other models, Gemini 4 Carbon and Argon, as best for deep reasoning on hard problems. Hydrogen is intended to be faster and cheaper for everyday multi-step work, so it can handle Search without using the most powerful model for everything.

More in Tech