Google AI Mode vs AI Overviews: What the citation gap means for your brand

Most marketing teams still treat Google's AI answers as one thing. They are two, and the difference now decides whether your brand shows up. AI Overviews are the summary box pushed to the top of a normal results page. AI Mode is a separate, conversational search experience users open on purpose, powered by Gemini and rebuilt again in November 2025 on Gemini 3. When Ahrefs compared 730,000 matched responses from both systems in September 2025, it found they agreed on what to say roughly 86% of the time, yet cited the same URLs only about 14% of the time. That gap is the whole story. Winning one does almost nothing to win the other, and the lever that moves both is not on your website.

AI Mode is a separate search engine, not a longer AI Overview

The first mistake to correct is treating AI Mode as AI Overviews with more words. They are distinct products with distinct behavior. AI Overviews launched broadly in the US in 2024 as a generated summary that appears above standard results and is pushed to you whether or not you asked. AI Mode is a dedicated, chat-style surface you pull open when you want to go deeper, and Google confirmed it moved the experience onto a custom Gemini model, then upgraded it to Gemini 3 on November 18, 2025 (Google). AI Overviews now reach roughly 2.5 billion users a month; AI Mode crossed about 1 billion within its first year.

Soar is a community marketing agency that has run 4,200+ community campaigns across 280+ brands since 2017, and the practical consequence we flag for clients is this: you are not optimizing for one Google AI answer, you are competing in two systems that pick sources differently. Treating them as one channel is how brands end up cited in the summary box, invisible in the conversation, and unable to explain why. For your board, the reframe is simple: AI Mode is a new search engine that happens to live inside Google.

Do AI Mode and AI Overviews cite the same sources?

Rarely, and this is the finding that should reset your strategy. Ahrefs compared AI Mode and AI Overviews on the same queries across 730,000 response pairs and 540,000 citation pairs from September 2025 US data. The two systems cited the same URLs only about 14% of the time, rising to just 16% when you compare only the top three citations in each answer. They shared an identical opening sentence 2.5% of the time. Yet semantic similarity averaged 86%, with 89% of pairs scoring above 0.8 on a scale where 1.0 is identical meaning (Search Engine Journal).

As the Ahrefs team put it, the two systems agreed on what to say and disagreed on whom to credit. That is not a rounding error you can ignore. It means a page that earns your brand a mention in AI Overviews has roughly a one-in-seven chance of also carrying you in AI Mode. If your agency reports "we are in Google's AI answers" without saying which one, the report is incomplete. The decision this forces: measure and pursue both surfaces separately, because success in one is not evidence of success in the other.

Across 730,000 matched answers, AI Mode and AI Overviews agreed on what to say 86% of the time. They cited the same sources only about 14% of the time.

How query fan-out multiplies the sources AI Mode reads

AI Mode does not run your search. It runs a research project. Google describes the underlying technique as query fan-out: the model breaks a single question into subtopics and issues many searches at once, then synthesizes one answer (Google). In practice a single prompt becomes 8 to 12 parallel sub-queries, and complex questions trigger far more. With Gemini 3, Google says fan-out runs even more searches and surfaces content it previously missed. iPullRank estimates that a keyword with 1,000 monthly searches can generate 8,000 to 12,000 retrieval events once fan-out is accounted for (iPullRank).

This changes the math of visibility in your favor if you understand it, and against you if you do not. Every sub-query is a separate chance to be cited, and the sub-queries are not the words your customer typed. A question like "best project management tool for agencies" fans out into pricing, integrations, team size fit, alternatives, and real-user complaints. If your brand only shows up on the literal head term and is absent from the comparison and complaint threads the fan-out pulls, you lose most of the retrieval passes. The takeaway for planning: you are not targeting a keyword, you are trying to be present across the dozen questions the model asks on your customer's behalf.

Google AI Mode vs AI Overviews: a side-by-side

The two surfaces differ on interaction, intent, and, critically, on which sources they reward. AI Overviews serve a fast, one-shot summary for a broad query. AI Mode serves a multi-step, conversational session for a researcher who wants to compare and drill down. Their citation behavior diverges accordingly, which is why a single "rank in Google's AI" plan underperforms. The table below is the version we put in front of clients when they ask why one report looks great and the other looks empty.

The row that matters is the last one. Because overlap is so low, a brand can dominate the summary box and be missing from the conversation, or the reverse. For a marketing leader, the implication is budget-shaped: you need a plan that produces the kind of source AI Mode's fan-out rewards, not just the tidy page an AI Overview might quote. We unpack the community mechanics behind this in how Google AI Overviews changed community marketing.

What actually earns citations in AI Mode

The sources AI systems cite are overwhelmingly off-page and community-driven, and this is consistent across every credible 2026 study. Semrush analyzed more than 230,000 prompts and 100 million citations across ChatGPT, Google AI Mode, and Perplexity and found Reddit the single most-cited domain overall, ahead of Wikipedia, LinkedIn, and Forbes (Semrush). A separate 30-million-source analysis reported by Search Engine Land ranked Reddit, YouTube, and LinkedIn as the top three cited sources across AI answers (Search Engine Land). Inside Google specifically, Reddit sits among the top sources in AI Overviews, and Quora ranks as the fourth most-cited domain in AI Mode.

The correlation data explains why. In Ahrefs' December 2025 study of 75,000 brands, branded web mentions correlated with AI visibility at roughly 0.66 to 0.71 across ChatGPT, AI Mode, and AI Overviews, while Domain Rating correlated at only 0.27 to 0.33. Off-page mentions are close to three times more predictive of AI citations than link authority. That is the strategic center of gravity: the fan-out reads what independent third parties say about you, and it trusts communities describing real experience over your own pages. We documented the origin of that trust in how Reddit became the biggest LLM citation source.

Why optimizing your pages does not transfer between the two

Here is the uncomfortable part for anyone whose AI plan is really an SEO plan. Ranking a page well in Google no longer predicts AI citation in either surface, and it definitely does not carry across both. Independent analyses put the overlap between AI citations and Google's own top 10 organic results at roughly 12%, and cited sources churn heavily, with 40% to 60% of cited URLs changing month to month across AI Mode and ChatGPT. A page tuned to rank can sit at position one and never appear in the AI answer sitting directly above it.

The reason ties back to fan-out and source selection. AI Mode is not looking for the best-ranked page on your head term; it is assembling an answer from the most trusted source on each of a dozen sub-questions, and freshness matters, with AI Mode showing a measurable preference for recently updated sources. So the same content investment produces uneven results: strong in the summary box, weak in the conversation, or the reverse, with no reliable way to predict which. This is why we tell clients to stop asking "how do we optimize our pages for AI" and start asking "where does the fan-out go, and are we there." The honest version of this diagnosis lives in backlinks vs brand mentions.

What this means for your budget and who should act now

The practical conclusion is clarifying rather than discouraging. One kind of work, earned presence in the communities and third-party sources AI trusts, pays off across AI Mode, AI Overviews, ChatGPT, Perplexity, and traditional search at the same time, because the credible source about your category is largely the same regardless of which engine is reading it. That is a rare efficiency: the fragmentation across surfaces is real for page optimization, and mostly absent for community presence. It typically runs $3,000 to $15,000 per month for an agency engagement, with a six-month minimum, because it compounds rather than spikes.

Who should prioritize this now: brands in categories where buyers research and compare in public before purchasing, and brands watching competitors get named in AI answers while they are absent. Who can wait: brands with no product-market fit, where forced community presence produces thin signals the model ignores, or low-consideration products buyers do not research. The click data raises the stakes either way. Pew found that when an AI summary appears, users click a traditional result only 8% of the time versus 15% without one, and click a link inside the summary just 1% of the time (Pew Research Center). Being the cited brand, not the ranked page, is increasingly the whole game. Start by finding out where you stand with a step-by-step AI visibility audit.

:::