A client came to us in June, quietly panicking. Their rankings had not moved. Their AI Overview citations looked healthy. And their organic traffic was falling off a cliff anyway, month after month, for no reason any of their dashboards could explain.
It took us one afternoon to find the culprit, and it was not on a single report they were watching. They were invisible in Google AI Mode, a search product that had crossed a billion monthly users while they were busy perfecting their strategy for the last version of search.
This is the article I wish that client had read in January. It is not a hot take. It is the manual. We are going to cover what AI Mode is, why the work you already did for AI Overviews barely counts here, the exact reasons it ignores most sites, and the playbook that gets you cited, with the 2026 data behind every claim and a test you can run in the next five minutes.
What Google AI Mode actually is
Google AI Mode is Google's conversational search experience, powered by Gemini. You ask a question in plain language, it writes you a full answer with citations, and then it remembers the thread so you can keep going. Ask a follow up and it builds on what it already told you.
Three things make it different from the search you know:
- It answers, it does not list. Instead of ten blue links, you get a written response assembled from several sources, with the citations tucked beside the claims.
- It is a conversation, not a query. The average AI Mode question runs 7.22 words versus 4.0 for classic Google, and follow-up questions are climbing more than 40 percent a month. People are talking to it, not searching it.
- It is already the main event, not a feature. AI Mode passed a billion monthly users about a year after launch, its query volume has more than doubled every quarter, and it now triggers on more than half of informational and commercial searches. Ninety-seven percent of its answers include at least one citation.
The takeaway is simple. AI Mode is not a box bolted onto search results. For a growing share of questions, it is becoming the search result, and it plays by rules most brands have never read.
The 13.7 percent problem
Here is the finding that should change how you spend your next quarter, and almost nobody is talking about it.
Google AI Mode and Google AI Overviews reach the same conclusion about 86 percent of the time, but they cite the same URL only 13.7 percent of the time. Same question, same answer, almost entirely different sources.
Read that again, because it quietly breaks the plan most teams are running. If you spent 2025 optimizing to get into AI Overviews, you optimized for roughly one in seven of the citations that now matter. AI Mode is a separate billion-user engine, sitting right next to the one you targeted, picking different pages by different rules.
The reason is something called query fan-out, and it is the single most important idea in this article. When you ask AI Mode a question, it does not run your question. It silently breaks your one question into dozens of smaller sub-questions, runs them in parallel, and assembles the final answer from what it finds, pulling from around seven unique domains per response. AI Overviews answers the one query you typed. AI Mode answers the twenty questions behind it. Different retrieval, different winners.
If your GEO plan was "get into AI Overviews," you have been optimizing for about one in seven of the citations that now decide who AI recommends. AI Mode is the other six, and it is bigger.
Seven reasons AI Mode ignores your site
When a site is invisible in AI Mode, it is almost always one of these seven reasons. Each one has a fix, and none of them require magic.
You optimized for AI Overviews and assumed it carried over
The fix. It does not. Treat AI Mode as its own channel with its own scoreboard. The content that wins is content that answers a whole topic, not a single query, because the engine is asking a whole topic's worth of sub-questions.
Your page answers the headline question but none of the follow ups
The fix. AI Mode fans one question into many. If your article nails "what is X" but says nothing about "how much does X cost," "is X worth it," "X versus Y," and "how do I start with X," you are eligible for one sub-query out of twenty. Cover the cluster, not the keyword.
Nobody searches for your brand by name
The fix. Brand search volume, not backlinks, is now the strongest predictor of whether AI recommends you. Models learn which names go with which categories from how often people look those names up. The unglamorous work of making people google you on purpose is now core AI visibility work.
You have no presence off your own website
The fix. Sixty-eight percent of AI citations come from third-party sources, and only thirty-two percent from brand-owned pages. Off-site authority such as Reddit, news, and reference sites is roughly 3.1 times more influential than on-page technical work. If the only place that vouches for you is you, the model has nothing to lean on.
Your best answers are buried in the middle of long paragraphs
The fix. A model lifts a passage, not a page. Every important answer needs to live in a self-contained chunk that makes sense pulled out of context, with the actual answer in the first sentence. Front-load the answer, then explain.
Your brand tells a different story in different places
The fix. The model cross-checks you against yourself. If your name, your one-line description, and your numbers differ between your site, your schema, and your profiles, you read as an unreliable narrator, and unreliable narrators get skipped. Make the entity identical everywhere.
Your content has not been touched in eighteen months
The fix. AI Mode weights freshness. A page that was last meaningfully updated in 2024 competes badly against one refreshed this quarter. Give your pillar pages a real update cadence and an honest dateModified, not a fake timestamp bump.
How AI Mode picks its sources
AI Mode does not pick sources the way classic search ranks pages, so it helps to know what is actually happening under the hood.
It starts by fanning out. Your 7-word conversational question becomes a fan of narrower queries. "What is the best CRM for a small agency" quietly becomes a dozen searches about pricing, integrations, ease of use, support, and specific competitor names.
It retrieves against each of those. For every sub-query it looks for the source that answers that exact slice best. This is why a page ranked eighth for the head term can still get cited: it might be the clearest answer to one of the sub-questions, even when the number one result is not.
It scores on four things. Semantic relevance to the sub-query, entity alignment (does the model already associate your brand with this topic), freshness, and structured data that makes your content machine-readable. Rank is an upstream filter, not the decision. Top-three pages are far more likely to be cited, but being cited is a separate contest you can win from further down the page.
It assembles from about seven domains. That means more citation slots per answer than a single AI Overview offers, which is good news: there is room for you, if your passages earn a slot.
The playbook that actually works
None of this is a trick. It is the same idea eight ways: be the most complete, most trustworthy, most machine-readable answer to the whole question, and prove it off your own website as much as on it.
Write for the fan-out, not the keyword
Stop building pages around one query. Build the definitive resource for a topic, the kind of comprehensive pillar that answers the head question and every sub-question a buyer would ask next. In practice that is long, genuinely thorough content, usually 2,500 to 5,000 words, structured so each sub-question has its own clearly labeled section. You are not writing for a keyword. You are trying to be the single page that answers the whole fan.
Make every passage stand on its own
Assume the model will lift one paragraph and show it with your name attached. Write so that paragraph survives the trip. Put the answer in the first sentence, keep the claim self-contained, and avoid pronouns that only make sense if you read the paragraph above. Clear, liftable passages get lifted.
Build brand search volume on purpose
This is the lever almost nobody pulls because it does not look like SEO. Get your name in front of people so they search for it: useful free tools, a point of view worth repeating, community presence, PR, partnerships, a product people tell their friends about. When search volume for your brand rises, the models learn that your name belongs in your category, and citations follow. It is slow, it compounds, and it is the closest thing to a moat in AI search.
Earn the third-party citations
Two thirds of AI citations live on sites you do not own, so go earn a place on them. The high-value ones in 2026 are the communities and reference sources models trust most: honest participation on Reddit and in niche forums, digital PR that lands you in real publications, original data studies journalists cite, and a clean, accurate entity presence on the reference sites. One credible outside mention outweighs a dozen self-serving pages.
Fix your entity so the model trusts it
Pick one name, one description, one set of facts, and use them everywhere: your site, your schema, your social profiles, your listings. Add Organization and Person schema with a sameAs graph that ties your profiles together. The goal is simple: when the model checks whether the different mentions of you agree, they should agree perfectly.
Give it the structured data it reads
Machine-readable content wins ties. Use Article schema on your posts, FAQPage schema on your answers, and HowTo schema on your step-by-step guides. Keep the important content in clean server-rendered HTML, because a meaningful share of AI crawlers do not run JavaScript, and content they cannot read is content they cannot cite.
Keep the pillars fresh
Set a real refresh cadence for your most important pages. Update the numbers, add the new sub-questions people started asking, and let the dateModified reflect genuine work. Freshness is a ranking and citation signal, and in a field that changes monthly, a stale flagship page is a slowly leaking one.
Measure citations, not just rankings
Your rank tracker will not tell you whether AI Mode names you. Track the thing that matters: for your top buying-intent questions, who does AI Mode actually cite. Watch that list move as you do the work. It is the only scoreboard that reflects the game you are now playing.
The five minute test you can run now
You do not have to guess whether any of this applies to you. You can see your exact standing in AI Mode in the time it takes to make coffee.
- Open Google AI Mode and start a fresh conversation, signed out if you can, so your own history does not skew it.
- Ask the exact question a real customer would ask, in full conversational sentences. Not "best CRM," but "what is the best CRM for a small marketing agency that hates complicated software."
- Read the answer and write down every domain it cited. There will usually be around seven.
- Ask yourself two questions: are you in that list, and is your closest competitor in it?
- Ask a natural follow up, the kind a real buyer would, and watch which sources stay cited as the conversation deepens. The ones that persist are the ones with topic-wide authority.
- Repeat for your five most important buying-intent questions. The set of domains that keep showing up is your real competitive set in AI search, and it is often not the one you thought you were fighting.
Most people are surprised twice: once by who gets cited instead of them, and once by how far down the page some of those cited competitors actually rank. That gap between rank and citation is the whole opportunity.
What a zero-click citation is worth
Let's deal with the objection head on, because it is the right one to raise. Between 92 and 94 percent of AI Mode sessions end without a click to an external site, compared to roughly 34 to 43 percent for classic Google search. If almost nobody clicks, why chase the citation at all?
Because the citation is the value, not the consolation prize. Being named in the answer is a brand impression delivered to a high-intent person at the exact moment they are deciding, and it compounds the brand-search effect that feeds the next answer. When people do click through from AI, they are worth more: AI-referred visitors spend around 15 minutes on site versus 8 from Google, and convert roughly three times better. Sites cited inside AI Mode answers see qualified traffic on those queries rise 20 to 35 percent, with conversion rates up to three times higher than normal organic clicks.
Zoom out and the stakes get concrete. McKinsey projects that 750 billion dollars in US revenue will flow through AI-assisted search by 2028, and warns that brands unprepared for it could lose 20 to 50 percent of their traditional search traffic as decisions move to AI before the click ever happens. Zero-click is not zero-value. It is pre-sold value going to whoever the model trusts.
One more thing, because you should know whether the people writing this actually do it. We run more than twenty of our own products, and across that portfolio we have earned over a million AI citations, measured in Bing over six months. One of those products, FreeCV, now gets more than half of its traffic from ChatGPT alone. Every tactic in this playbook, we ran on ourselves before we sold it to anyone.
The pre-flight checklist
Before you decide you are done, run down this list. If you can honestly check every box, you are doing the work that gets a brand cited in AI Mode in 2026.
Keep reading
How to Rank in Google AI Overviews
The other half of Google AI search, and why the same work does not transfer.
FoundationWhat Is GEO?
The bigger discipline of getting cited by AI, defined properly.
Related guideHow to Get Cited by ChatGPT
The ChatGPT side of the same game. Different engine, related signals.
Our serviceAI Mode Optimization
Want us to run this playbook for you? This is the service that does it.
Want to be the brand AI Mode names?
We earned over a million AI citations for our own products before we sold this to anyone. Let us do it for yours. Start with a free audit and we will show you exactly where you stand in AI Mode today.
Get a Free Audit
Abd runs Generative Engine Optimization across the Outline portfolio, where more than twenty products have earned over a million AI citations between them. He writes about what actually moves AI visibility, tested on real products first.
FAQ: real questions answered
Q.What is Google AI Mode?
Google AI Mode is Google's conversational, Gemini-powered search experience. Instead of a list of links, it writes a full answer with citations and remembers the conversation so you can ask follow ups. It passed a billion monthly users about a year after launch and now triggers on more than half of informational and commercial searches.
Q.How is AI Mode different from AI Overviews?
AI Overviews is a summary box that sits above the links for a single query. AI Mode is a full conversational session that breaks your question into many sub-questions and answers all of them. The practical difference is stark: the two features reach the same conclusion about 86 percent of the time but cite the same URL only 13.7 percent of the time, so optimizing for one does very little for the other.
Q.How do I get cited in Google AI Mode?
Build comprehensive content that answers a whole topic and its sub-questions, write self-contained passages with the answer in the first sentence, grow your brand search volume, earn third-party citations on the sources models trust, keep your entity consistent everywhere, add structured data, and keep your pillar pages fresh. Rank helps but does not decide, because a lower-ranked page can still be the best answer to one of the sub-questions.
Q.Does AI Mode hurt my traffic?
It changes it. Between 92 and 94 percent of AI Mode sessions are zero-click, so raw visit counts on informational queries fall. But being cited is still valuable, and the clicks you do get are higher intent, spending more time on site and converting several times better. The brands that lose are the ones that are neither cited nor clicked.
Q.Can I optimize for AI Mode and AI Overviews at the same time?
The fundamentals overlap: clean content, clear passages, entity authority, and structured data help in both. But because the two features cite the same URL only 13.7 percent of the time, you should treat them as related but separate goals and measure each on its own. Winning one does not automatically win the other.
Q.Does traditional ranking still matter for AI Mode?
Yes, as an upstream filter. Top-three pages are far more likely to be cited, so strong organic rankings still help you qualify. But citation is a separate contest. A page ranked eighth can be cited when it answers a specific sub-question better than the top result, so rank is necessary insurance, not a guarantee.
Q.What is query fan-out?
Query fan-out is the technique AI Mode uses to answer a question. It silently breaks your single question into dozens of narrower sub-questions, runs them in parallel, and assembles the final answer from the best sources for each. It is why comprehensive content that covers a whole topic beats a page built for one keyword.
Q.How many sources does AI Mode cite per answer?
Around seven unique domains per response, and 97 percent of AI Mode answers include at least one citation. That is more citation slots than a single AI Overview offers, which means there is genuine room to earn a place if your passages are strong.
Q.Why does AI Mode cite my competitor and not me?
Usually one of three reasons: your competitor has more brand search volume, so the model associates their name with the category more strongly; they have more third-party citations on the sources models trust; or their content answers the specific sub-question better than yours. Run the five minute test, see who gets cited for your queries, and close whichever gap is biggest.
Q.Is brand search volume really more important than backlinks now?
For AI citations, the 2026 data says yes. Brand search volume has become the single strongest predictor of whether AI recommends you, and off-site authority is roughly 3.1 times more influential than on-page technical work. Backlinks still matter for classic SEO, but if you want to be named by AI, making people search for you by name is the higher-leverage move.
Q.How long until I show up in AI Mode?
If you already have authority and rankings on a topic, improving a page can start earning citations within weeks, because AI Mode recrawls and reassembles constantly. If you are starting from a weak position, expect a few months of building brand search, third-party citations, and topic depth before AI Mode treats you as a trusted source. There is no instant path, and anyone selling one is selling a scam.
Q.Do I need structured data to be cited in AI Mode?
It is not strictly required, but it is a meaningful advantage and it wins ties. Article, FAQPage, and HowTo schema make your content easier for the model to parse and reuse, and clean server-rendered HTML makes sure the many crawlers that do not run JavaScript can actually read you. It is low effort for a real signal.
Q.How do I track AI Mode citations?
Start manually: run your top buying-intent questions through AI Mode on a schedule and record which domains get cited. For scale, specialized AI visibility tools now monitor citations across AI Mode, AI Overviews, ChatGPT, and Perplexity. The point is to track who the model names for your queries, not just where you rank.
Q.Does AI Mode optimization work for local businesses?
Yes, with a local twist. AI Mode leans on the same trust signals plus local ones: a complete and accurate Google Business Profile, real reviews, consistent name, address, and phone details everywhere, and detailed service-area content. Ask AI Mode a near-me question in your category and see who it names, then close the gaps.
Q.Is optimizing for AI Mode worth it if most sessions are zero-click?
Yes. The citation itself is a high-intent brand impression at the moment of decision, the clicks you do earn are worth several times more than average, and the whole thing compounds your brand search, which feeds future citations. With hundreds of billions in revenue projected to move through AI-assisted search, being the name the model trusts is worth far more than a raw visit count.