Gemini powers two separate Google AI surfaces that most brands are tracking as one thing. AI Overviews appear above traditional organic results and still show blue links below. AI Mode replaces the results page entirely with a Gemini-powered conversation. They use the same underlying model but retrieve differently, surface in different query contexts, and require slightly different optimisation approaches.
Together they appear on over 55% of all Google searches. Google AI Overviews reach 2 billion monthly users. Google AI Mode crossed 1 billion monthly users before Google I/O 2026 in May and became the global default search experience at that event. Getting cited by Gemini is now the largest single AI search opportunity available to most brands, by volume.
But Gemini AEO has a specific characteristic that separates it from Perplexity or ChatGPT optimisation: it draws from Google's standard index. That means your existing SEO infrastructure either helps you or hurts you directly. There is no separate Gemini retrieval system to optimise for independently — you are optimising the same pages that Googlebot crawls, in the same index that produces traditional search rankings.
How Is Gemini Different From Perplexity and ChatGPT as a Retrieval System?
Perplexity retrieves in real time on every query. ChatGPT uses a mix of model memory and live retrieval. Gemini — powering both AI Overviews and AI Mode — draws from Google's standard Search index, the same one that produces your organic rankings.
This has three practical implications:
First, if a page is not indexed by Google, Gemini cannot cite it. Canonical errors, noindex tags, duplicate content filters, and poor crawl budget allocation all block Gemini citations in the same way they block organic rankings. Standard technical SEO is prerequisite, not supplementary.
Second, Gemini can render JavaScript. It is the only major AI search engine that fully executes JavaScript before indexing, the same way Chrome renders a page. This means client-side rendered content — React SPAs, Next.js without SSR — is accessible to Gemini in a way it is not accessible to PerplexityBot or OAI-SearchBot. Your Gemini citation potential is not limited by JavaScript rendering the way your Perplexity potential is.
Third, Gemini inherits Google's E-E-A-T weighting. The same quality signals Google uses for organic ranking — expertise, authoritativeness, trustworthiness, experience — feed directly into Gemini's citation selection. A page that ranks well in organic search because it has strong E-E-A-T signals is a better Gemini citation candidate than a structurally equivalent page with weak E-E-A-T. The channels reinforce each other.
What Does Gemini Prioritise for AI Overview Citations?
AI Overviews appear for queries where Google's systems judge that a synthesised answer will serve the user better than a list of links. They are most common on informational and research queries. The citation selection criteria for AI Overviews:
Passage extractability at the top of the page. Google's passage indexing algorithm scores individual passages rather than full pages. Passages in the first 30% of a page are scored with higher retrieval weight. A direct, self-contained answer in the first paragraph of each H2 section is the single highest-impact structural change for AI Overview citations. SparkToro's 2026 data confirmed 44.2% of all AI citations come from the first 30% of content — this pattern is especially pronounced in Google AI Overviews.
FAQPage schema. Google removed FAQPage rich results from standard search in May 2026 but explicitly confirmed FAQPage JSON-LD continues to be parsed for AI Overview and AI Mode retrieval. Each question-answer pair in the schema is a machine-readable extraction target. Do not remove FAQPage schema from pages that already have it, and add it to any page with question-and-answer content.
Fresh dateModified signals. AI Overviews weight freshness. A page updated last week beats a structurally identical page last updated 18 months ago. Update dateModified in your Article JSON-LD every time you make material content changes. Update lastmod in your XML sitemap at the same time. Submit updated URLs via Google Search Console URL Inspection to accelerate re-crawl.
Core Web Vitals — especially LCP. Google's crawl budget allocation prioritises pages with strong Core Web Vitals. Pages with LCP above 4 seconds get crawled less completely. Incomplete crawls produce incomplete passage extraction. LCP under 2.5 seconds is the threshold that consistently correlates with full passage indexing in NotionCue's data across 8,000 tracked domains.
What Does Gemini Prioritise Differently in AI Mode?
AI Mode uses the same Gemini model but runs on a different retrieval architecture from AI Overviews. The key difference is the fan-out technique, documented in the Google AI Mode technical guide: AI Mode issues up to 16 sub-queries for a single user question, retrieving the best available source for each sub-query separately.
This changes the citation opportunity. A page that does not match the primary user query can still be cited if it is the best available source for one of the sub-queries AI Mode generates internally. A page about "how to track AI citations" can appear in AI Mode answers to "how do I improve my brand's AI search visibility" because tracking is a sub-component of the broader improvement question.
For AI Mode specifically, topical completeness matters more than exact keyword match. A page that covers its topic in depth — answering the primary question plus three or four related sub-questions — provides more fan-out citation opportunities than a page that answers only the primary question at similar quality.
The content architecture that works for AI Mode is the pillar-cluster structure: a comprehensive pillar page linking to six or more spoke pages each covering a specific sub-topic. Each spoke is a separate fan-out candidate. The NotionCue blog series itself is an example of this structure — each technical post is a spoke that can surface in AI Mode answers to broader AEO questions.
What Schema Does Gemini Specifically Respond To?
Google explicitly names structured data as a supporting signal for AI features in its May 2026 official guide. The schema types that most directly affect Gemini citation probability, in priority order:
- FAQPage JSON-LD. Highest single impact. Creates directly injectable Q&A pairs for AI retrieval. Apply to any page with question-and-answer content.
- Article with datePublished and dateModified. Freshness signal. Update dateModified on every material content change. This field is more important than most teams realise — it is how Gemini determines whether your content is current.
- Organisation with sameAs. Entity authority. Links your brand to LinkedIn, Crunchbase, Wikidata, and other profiles that feed Google's Knowledge Graph. The same Knowledge Graph that powers traditional search Knowledge Panels feeds Gemini's entity confidence scores.
- BreadcrumbList. Topical hierarchy signal. A page about "AEO prompt tracking strategy" sitting inside a structured AEO content cluster carries more topical authority than a standalone page on the same topic.
- Person schema on author pages, linked from Article schema. E-E-A-T author signal. See the E-E-A-T and AI citation guide for the full implementation.
How Do You Track Gemini Citations Separately From Traditional Search?
Google Search Console added AI Overview and AI Mode reporting in 2026, but does not yet allow filtering to separate AI surface impressions from standard organic impressions cleanly. The data is mixed.
Three practical tracking approaches for Gemini-specific performance:
Question-format query filtering in GSC. Filter your Search Console queries to those starting with what, how, why, which, when. These are the query types most likely to trigger AI Mode and AI Overviews. If impressions for these queries rise but clicks fall, AI Mode is absorbing the intent. That pattern tells you Gemini is relevant for those queries and that appearing inside the AI response — not ranking below it — is the metric that matters.
Impression-to-click ratio trend. Track this ratio monthly for your top 20 informational queries. A declining ratio means more impressions are being absorbed by AI surfaces before generating clicks. A stable ratio means AI surfaces are not significantly affecting that query set. This distinguishes which query clusters need AEO attention versus which are not significantly impacted yet.
Manual AI Mode prompt runs. Run your 15 most important tracked prompts directly in AI Mode (separate from AI Overviews) once per week. Record whether your brand appears, what source is cited, and which competitor appears when you do not. This is the ground-truth data that GSC cannot currently provide. The NotionCue Prompt Tracker automates this tracking at scale.
Gemini draws from Google's standard index, so the NotionCue AI Crawler Audit checks Googlebot access specifically for AI surface indexing alongside the other AI crawler user-agents. Pages with crawl issues that affect traditional rankings have the same issues in Gemini. Fix them once and both channels benefit — the efficiency advantage of Gemini AEO versus optimising for Perplexity or ChatGPT separately.
Frequently Asked Questions
Do AI Overviews and AI Mode cite the same pages?
No. SLIDEFACTORY's June 2026 analysis found only 14% URL overlap between AI Mode citations and AI Overview citations. They use the same Gemini model but different retrieval architectures. AI Overviews pull from a more curated, authority-weighted candidate set. AI Mode's fan-out technique retrieves from a broader set of sub-query candidates. Track them separately in your prompt monitoring.
Does Google's May 2026 guide say anything different from what practitioners have been doing?
It confirms the approach rather than changing it. The guide states that AI Overviews and AI Mode draw from Google's standard index, that structured data supports AI visibility, and that there is no separate optimisation required for AI surfaces beyond good foundational SEO plus structured data. The one surprising addition: it explicitly lists llms.txt as a tactic Google does not use for AI Overview or AI Mode eligibility.
Is there anything unique about Gemini that ChatGPT or Perplexity does not do?
Gemini is the only major AI search engine that fully renders JavaScript before indexing — the same way Chrome processes a page. This means Gemini can see content that is invisible to PerplexityBot and OAI-SearchBot. For brands with client-side rendered sites, Gemini citation potential is higher than Perplexity citation potential on equivalent pages. Check Perplexity access via curl to confirm which content is invisible to non-Gemini engines.
If I am already ranking top 3 on Google, how much extra AEO work do I need for Gemini?
Less than for Perplexity or ChatGPT. Strong organic rankings indicate your pages are indexed, crawlable, and trusted by Google — which directly feeds Gemini. The gap is in passage extractability (BLUF structure) and FAQPage schema. Add those two elements to your top-ranking pages and Gemini citation rates improve without additional infrastructure work.