Google AI Mode crossed 1 billion monthly users with Gemini 3.5 Flash as the default. Top-10 rankings now drive only 38% of AI Mode citations, down from 76%. Here is what changed and what to fix.
By Stephan Charles | Last fact-checked: <time datetime="2026-06-24">June 24, 2026</time>
Google AI Mode reached 1 billion monthly users in May 2026 and now runs on Gemini 3.5 Flash as the default model across 200 countries and 98 languages. The model change shifts citation patterns in a measurable way: top-10 Google results accounted for 76% of AI Overview citations in mid-2025 and account for about 38% in early 2026, per analysis of AI Mode citation shifts. Domain authority no longer predicts which passages get extracted and cited. Structured content, FAQ schema, and answer-first passage design now determine whether Gemini 3.5 Flash cites your brand. BrandCited is an AI brand visibility platform that monitors citation rates across 9 AI engines: ChatGPT, Perplexity, Claude, Gemini, Copilot, Grok, You.com, Brave, and Google AI Mode. BrandCited scores each brand on a 0-to-100 AI visibility index. For a brand whose SEO team has been tracking rankings but not citation structure, the Gemini 3.5 Flash upgrade is the point where the gap becomes measurable.
BrandCited's free scan checks your AI Mode citation rate alongside 8 other engines in 30 seconds and shows exactly which queries your brand answers and which it misses.
See exactly how AI engines like ChatGPT, Perplexity, and Gemini perceive your brand.
Start free scanStephan Ochse
Industry News
The BrandCited team covers GEO, AI search optimization, and brand visibility strategy. We publish research, practical guides, and product updates every week.
AI referral traffic converts at 14.2% versus Google organic's 2.8%. Perplexity Comet Plus now pays publishers 80% of citation revenue. Here's what the data means for brand visibility.
16 min read
G2: 51% of B2B buyers start vendor research in AI chatbots, 69% chose a vendor they found there. What this means for your brand's AI citation strategy.
9 min read
Google Search Console added generative AI performance reports on June 3, 2026: impressions only, UK-first, Google features only. ChatGPT, Perplexity, and 6 other engines are not covered.
10 min read
<section id="gemini-flash-citation-changes">
Gemini 3.5 Flash pulls sources from web search without an explicit browse request from users, which lowers the bar for inclusion as a cited source — but only for content structured as atomic, extractable units, per Nettpilot's analysis of Gemini 3.5 Flash and content visibility changes in 2026.
Atomic fact 1: Gemini 3.5 Flash is the default model inside Google AI Mode for all users worldwide, deployed across 200 countries and 98 languages at no subscription cost, per Google's official Gemini 3.5 Flash model page.
Atomic fact 2: Google AI Mode has surpassed 1 billion monthly users, with queries more than doubling every quarter since launch, per Google's I/O 2026 Search announcement.
Before the Gemini 3.5 Flash upgrade, citations in AI Mode required a user to select browse mode or ask for sources. Gemini 3.5 Flash defaults to source-backed responses for factual queries. That means a brand now needs to be in Google's live index AND have passage-structured content — two requirements that didn't both apply before.
The scale matters. AI Mode processes more monthly queries than many search engines handle in a year. A brand that isn't structured for extraction isn't eligible for those citations, regardless of its organic ranking position.
</section>
<section id="ai-mode-citation-shift">
AI Mode citation selection uses a two-step process: Google's retrieval layer finds candidate pages through its standard index, then Gemini 3.5 Flash selects which passages to extract and cite based on content structure. The second step uses different signals from PageRank.
Atomic fact 1: Top-10 Google results accounted for 76% of AI Overview citations in mid-2025 and account for about 38% in early 2026, per analysis of AI Mode citation patterns from Influencers Time. Pages deeper in organic rankings win citations over top-ranked pages that lack structured passage design.
Atomic fact 2: Passage structure and information consistency drive citation selection — ranking well supports indexability but doesn't determine which passages AI systems extract and synthesize, per Discovered Labs' analysis of the May 2026 search update.
The mechanism is clear. Google's ranking algorithm scores pages for relevance based on domain authority, link equity, content freshness, and semantic relevance. Gemini 3.5 Flash, after retrieval, scores passages for extractability based on self-contained meaning, named entities, and answer-first structure. A page that ranks first but buries the answer in the third paragraph of the fifth section loses citations to a page that ranks third but opens each H2 section with a direct answer.
BrandCited's scan data shows 43% of brands with a top-5 Google position for their primary keyword score below 30 on the AI visibility index. They appear in fewer than 30% of AI-generated answers for their target queries across BrandCited's 9-engine panel.
</section>
<ScanWidget />
<section id="content-structure-gemini-flash">
Gemini 3.5 Flash cites content that forms a citable unit at the passage level: 200-to-400-word sections that open with the answer in the first sentence, each containing at least one named entity and one specific number, per ALM Corp's analysis of Google AI Mode citation patterns in 2026.
Atomic fact 1: Content with statistics, citations, and quotations achieves 30-to-40% higher visibility in AI responses, and pages updated within 2 months earn 28% more citations than older content, per Onely's 2026 analysis of AI search visibility factors.
Atomic fact 2: AI-referred visitors spend 68% more time on websites than visitors from traditional organic search, per Superlines' 2026 AI Search Statistics report. The citation audience is the highest-engagement traffic segment on site.
The structural requirements for Gemini 3.5 Flash citations map to the Schema.org Article type, which tells AI engines the author, publisher, publication date, and primary topic of each page. Article schema is one of three schema types BrandCited's audit flags as P0 — the highest-priority fix for brands with low AI visibility scores.
Question-formatted H2 headings are the single most impactful structural change for most brand pages. An H2 like "What is [your product category]?" signals to Gemini 3.5 Flash that this section answers a question — the content type AI Mode surfaces in response to user queries. An H2 like "Benefits of [your product]" signals a features-and-benefits narrative, which AI Mode filters out in favor of answer-format content.
</section>
<section id="schema-types-ai-mode">
FAQ schema (FAQPage JSON-LD) achieves a 67% citation rate in structured AI responses and is the single highest-impact structured data change for AI Mode citation, per SEO Score Tools' 2026 analysis of FAQ schema performance. Pages with 3-to-4 complementary schema types earn citations at 2x the rate of pages with a single schema type.
Atomic fact 1: Content with proper schema markup has a 2.5x higher probability of appearing in AI-generated answers than equivalent content without schema markup, per Stackmatix's 2026 structured data and AI search guide.
Atomic fact 2: Google's official AI optimization guide for developers states that generative AI features on Google Search pull from the same signals that determine search quality — structured data, author clarity, and content freshness are all named inputs in Google's public documentation.
The three schema types that combine for the strongest AI Mode citation signal: FAQPage for Q&A content, Article for named author and publisher, and BreadcrumbList for site structure context. BreadcrumbList helps AI engines understand where a page sits in your site hierarchy, which affects the topical depth signal. A blog post at /blog/ai-search/faq-schema/ carries a clearer topical signal than one at /blog/post-47.
The Schema.org FAQPage specification defines the JSON-LD format. Five to eight questions per page, each with a 2-to-4-sentence answer. Questions should match what users type into AI engines — not the questions your brand wants to answer.
BrandCited's AI Mode audit scores schema coverage as a composite: Article schema present, FAQPage schema present, BreadcrumbList present, and schema nesting valid. A brand missing all three on its core pages shows a near-zero AI Mode citation rate regardless of organic ranking.
</section>
<section id="ai-search-updates-last-24-hours">
Track your AI visibility for free
See how ChatGPT, Claude, Gemini, and 4 other AI platforms mention your brand.
</section>
<section id="how-brandcited-audits">
BrandCited's audit engine checks AI Mode citation rates as part of its Gemini engine score, one of 9 engine-specific scores in the composite AI visibility index. The audit includes five Gemini-specific checks: Article schema present, FAQPage schema present, BreadcrumbList present, answer-first H2 structure on core pages, and passage self-containment (no pronoun dependencies across sections). Run a free BrandCited scan at brandcited.ai to see your current AI Mode citation rate alongside your score across all 9 engines.
</section>
BrandCited scans your brand across 9 AI engines including Google AI Mode and shows your AI visibility score alongside citation gaps by engine, query type, and content page. The scan takes 30 seconds and requires no signup. Run a free AI visibility audit at brandcited.ai.
<section id="what-to-do-right-now">
Take these five actions in order of impact on AI Mode citation rate.
</section>
<section id="frequently-asked-questions">
Does Google's AI Mode use different content than regular Google Search for citations?
AI Mode's citation layer uses Gemini 3.5 Flash to select which passages to extract from the pages Google retrieves. Standard Search ranks pages based on authority and relevance signals. AI Mode adds a second filter: passage extractability. A page that ranks on page one but buries answers in narrative prose loses citations to pages that rank lower but open each section with a direct answer.
How many monthly users does Google AI Mode have?
Google announced at I/O 2026 in May that AI Mode has surpassed 1 billion monthly users, with queries more than doubling every quarter since launch. The rollout covers 200 countries and 98 languages, with Gemini 3.5 Flash as the default model worldwide.
What's the difference between Google AI Overviews and AI Mode?
Google AI Overviews appear at the top of standard Search results for specific query types. AI Mode is a full conversational interface, now available to all users worldwide without a subscription. Both use Gemini models for generation, but AI Mode handles longer, multi-turn queries and pulls sources from a broader range. BrandCited tracks citation rates in both surfaces separately.
Why does my brand rank on page one but not appear in AI Mode answers?
The two surfaces use different selection criteria. Google Search ranks your page based on authority, relevance, and PageRank signals. AI Mode selects passages based on structure: answer-first openings, named entities in the lead sentence, FAQ schema, and passage self-containment. A brand can hold a page-one ranking and still score near zero on BrandCited's AI Mode citation check if its content uses narrative-essay structure without schema markup.
How long does it take for content changes to improve AI Mode citation rates?
For retrieval-augmented engines including Google AI Mode, Perplexity, and Copilot, content changes appear in citation behavior within 7 to 30 days of crawl — faster if you submit the updated page to Google Search Console. Schema additions are the fastest path: FAQ schema changes can affect AI Mode citation behavior within 14 days of implementation and indexing.
What AI visibility score does BrandCited give brands?
BrandCited's AI visibility score is a 0-to-100 composite index that measures how often a brand appears in AI-generated answers for its target queries, across 9 engines: ChatGPT, Perplexity, Claude, Gemini, Copilot, Grok, You.com, Brave, and Google AI Mode. Each engine contributes an engine-specific sub-score. The composite score weights engines by query volume and citation architecture.
</section>
<!-- cite-widget: apa | mla | chicago | bibtex | permalink -->