Insight · October 16, 2026
One question,
a dozen searches.
Google AI Mode does not rank ten links. It breaks your question apart, searches each piece at once, and writes one answer. That changes what being found means.
01 · The short answer
What Google AI Mode is.
Google AI Mode is Google's conversational AI search experience, a dedicated mode where a custom version of Gemini answers a question directly instead of returning a list of links. It breaks the question into subtopics, runs many searches at once across Google's index, and writes one synthesized, cited answer you can keep asking questions about.
Google introduced it in 2025 and moved it out of the labs fast. By May 2026 the company said AI Mode had surpassed one billion monthly users, with its queries more than doubling every quarter since launch. It is no longer an experiment at the edge of Search. For a growing share of questions, it is Search.
02 · Not the same as AI Overviews
One sits on the results page. The other replaces it.
People use the two names as if they mean the same thing. They do not. An AI Overview is a summary block that sits at the top of the normal results page, above the links you already know. You still get the page. The overview is a shortcut on it.
AI Mode is a separate destination. You ask, and the page you get back is an answer, not a ranked list. You can ask the next question without starting over, bring in an image, or lean on live sources the model pulls at the moment you ask. Google has said the two are converging into one AI Search experience, which only raises the stakes of understanding how the deeper one works.
The one line to keep
“You cannot optimize the page anymore. You can only be worth citing inside it.”
03 · Query fan-out
How one question becomes one answer.
Google's Head of Search, Elizabeth Reid, described the method at Google I/O 2025. Search calls on a custom version of Gemini to break a question into different subtopics and issue a multitude of queries at the same time. Five beats, under the hood.
01
Read intent
Search recognizes when a question needs reasoning rather than a list of ten blue links, and hands it to a custom version of Gemini.
02
Break it apart
The model splits one question into subtopics. A single query about a trip becomes separate questions about weather, cost, timing, and more.
03
Fan out
It issues a multitude of searches at once against Google’s live index, so classic ranking still decides what each branch can see.
04
Synthesize
The model reads across the returned results, reconciles them, and writes one answer rather than handing you a page to assemble yourself.
05
Cite and continue
It shows links for the claims it leans on, and stays open to the next question so the thread keeps its context.
04 · What changed in 2026
Faster underneath, and the page itself is now generated.
The reach grew first. After a US, UK, and India start, Google expanded AI Mode to more than 180 countries in English in August 2025, then added languages and more territories through the autumn. At Google I/O 2026 the company made Gemini 3.5 Flash the default model behind it.
The bigger shift is the page. Google introduced generative UI, where Search assembles a custom layout for the question with interactive visuals, tables, graphs or simulations built in real time. The result is no longer a fixed template with your link somewhere on it. It is a layout the model draws for each question, and your content is raw material it may or may not reach for.
05 · One question, pulled apart
What the fan-out looks like from the inside.
Say a founder asks AI Mode which design partner to hire for an AI product launch. A classic search returns a ranked list and leaves the founder to read, compare, and decide. AI Mode does the pulling apart itself. The one question becomes a set of quieter searches, each run against the live index, and the model reconciles what comes back into a single answer.
You never see the branches. The brand that gets named is not the one that ranked first for the head term. It is the one that kept surfacing across the subtopics, with content a model could read and trust on each of them.
QUESTION "best design partner for an AI product launch"
FAN-OUT → what does an AI product design partner do
→ agency vs in-house for an AI launch
→ how to evaluate an AI design studio
→ typical AI product design engagement
→ signs of a good AI product team
SYNTHESIS one answer, drawn from the sources that held
up across every branch aboveIllustration of the method · the real subtopics are chosen by the model
06 · What earns visibility now
Stop chasing the rank. Start being the source.
When one question becomes a dozen searches, ranking first for the head term stops being the goal. The goal is to be the source the model keeps reaching for no matter which subtopic it is on. That rewards depth across a topic, not a single page tuned for a single phrase.
Three things move it. Write answer-shaped content, a clear claim first, structured so a model can lift it cleanly. Be a clear entity, an unambiguous thing a model can name and trust. And keep the technical floor solid, because the fan-out runs on the Search index, so classic ranking still decides what each branch can even see. Google's own guidance for AI search has not changed either: original, useful content made for people, not bulk pages made for a machine.
This is the work behind AI visibility, where a brand gets a free score for where it shows up in ChatGPT and AI search from live answer engines, then the playbook to raise it. It sits alongside ranking in AI Overviews and how Gemini chooses sources.
07 · What it does not change
The web did not disappear. It got read differently.
AI Mode still runs on the open web. The fan-out searches the same index the links come from, which means the content you publish is what the model reads. When Search does not have high enough confidence in the quality of an answer, Google says AI Mode falls back to a set of web links, the familiar page underneath.
So the fundamentals hold. Clear, accurate, well structured pages are still the asset. What changed is the reader. For years you wrote for a person scanning a results page. Now you are also writing for a model that will read ten of your pages at once, quote the parts that hold up, and never show the person the list you used to compete on.
Closing
The page is now generated per question.
Open AI Mode. Ask it the question your customers ask, the one that should bring them to you. Watch which brands it names and which it leaves out. That answer, the one you cannot rank your way into, is the brief for the next year of your content.
Written by · October 16, 2026
Donny Smith · ECD, Founder, Bttr.
Over the past 15 years, he has led creative teams and contributed to products used by millions of people worldwide, working with companies including Apple, Opendoor, JP Morgan, GE Aerospace, Pepsi, and Alterra Mountain Company.
LinkedInGoogle · Expanding AI Overviews and introducing AI Mode · May 2025 · Google · AI Mode in Google Search, updates from Google I/O 2025 · May 2025 · Google · 100 things we announced at I/O 2026 · May 2026 · Google · Get AI-powered responses with AI Mode in Google Search · Engadget · Google AI Mode is expanding to 180 countries · August 2025 · Aleyda Solis · Google's query fan-out technique · 2025 · Facts current as of October 16, 2026
Share this perspective
More insights
Adjacent perspectives.
Bttr. Field Brief
The brief Bttr. writes for senior buyers.
Monthly. One signal worth your time on Brand Operating Systems, AI search visibility, and the infrastructure buildout. No filler.