Get cited by Gemini.

Two surfaces, one foundation.

Gemini's optimization story has more moving pieces than any other GEO platform — because it spans Google's full retrieval graph plus a frontier-model generation layer.

Gemini models. The frontier model family that powers all Gemini surfaces — Gemini 2.5 Pro, Flash, and successors. Trained on Google's licensed corpora, Common Crawl, code, and Google's own product data. Model cards are public; specific training data is not.

Google Search index. The retrieval substrate for AI Overviews and most Gemini-app web-search queries. The same index that produces the blue links produces the candidate pool Gemini grounds in. Ranking on Google is the eligibility gate.

AI Overviews. The Gemini-generated answer cards that appear above the blue links for an increasing share of Google queries. AI Overviews quote multiple sources inline, with citation links to the underlying pages. Eligibility is driven by ranking, structure, and intent fit.

Gemini app + Workspace. The conversational surface and the Workspace integrations (Docs, Gmail, Sheets) that route 8M+ enterprise seats through Gemini for daily work. Citation behavior on the conversational app is closer to Perplexity than to AI Overviews — explicit inline links, source attribution.

Google-Extended. The robots.txt token that controls whether Google can use your content to train Gemini. Allowing it doesn't change SEO ranking; blocking it doesn't either. But blocking it does remove your content from Gemini's training set — affecting closed-book recall on the next training cycle.

Winning the answer card above the blue links.

AI Overviews are the highest-leverage AI-search surface on the open web — exposed to 2 billion monthly users inside Google Search, and increasingly the only surface searchers ever read. For an expanding share of queries, the AI Overview is the search result. The blue links below it are decoration. Optimization for AI Overview citation is the most consequential GEO work most brands can do.

Eligibility starts with rank. Pages that don't rank in the top 10-20 organic results are rarely cited in AI Overviews. The candidate pool is drawn from the same retrieval substrate as the blue links. Strong SEO is a prerequisite, not an alternative.

Then it filters on extractability. Among ranking pages, AI Overviews cite the ones whose content is structurally easiest to chunk and quote: pages with direct-answer leads, FAQ schema, ordered lists, comparison tables, and clear heading hierarchy. The same patterns that win on Perplexity win in AI Overviews — for the same retrieval-augmented reasons.

Intent match matters more than keyword density. AI Overviews are reranked against the underlying intent of the query, not the literal keyword. Pages that comprehensively answer the intent — even with looser keyword match — are picked over pages that match keywords precisely but answer adjacent questions.

SGE → AI Overviews evolution. AI Overviews evolved from Google's earlier Search Generative Experience (SGE) experiment. The signal patterns have stabilized over multiple iterations, and we've tracked them across the transition. The work that earns AI Overview citations today is the work that's compounded across two years of generative-search rollout.

Bot access, schema, and the Google-Extended decision.

Gemini's technical layer is mostly Google's technical layer — Googlebot crawls, Google's index serves, Google's signals weigh. The distinctive GEO-specific decisions concern Google-Extended and AI Overview-specific schema.

Googlebot. Same crawler, same crawl-budget rules, same render-engine that's already producing your organic rank. AI Overview eligibility inherits all of it. Technical SEO debt limits AI Overview visibility the same way it limits blue-link visibility.

Google-Extended. A separate robots.txt token (added October 2023) that controls whether Google may use your content to train Gemini. Disallowing it doesn't affect Search ranking or AI Overview citation; it only removes your content from training. Most clients allow Google-Extended deliberately so Gemini's closed-book recall includes their brand. A few publishers block it as a content-licensing stance.

Schema. JSON-LD has unusually high leverage on Gemini surfaces. AI Overviews appear to use schema as the primary structural signal for whether a page is fit-to-cite. Article, FAQPage, Product, Organization, BreadcrumbList — all foundational. We deploy Q&A and HowTo schema where appropriate as additional AI-Overview accelerators.

Core Web Vitals. Same thresholds as classic SEO — LCP, INP, CLS — apply. Pages in the green tier are cited more often than pages in the red tier when other signals are tied.

llms.txt. Google hasn't formally adopted llms.txt, but other AI surfaces that retrieve Google-indexed content (e.g., third-party assistants pulling Google search results) increasingly do. We deploy llms.txt on every Gemini engagement to keep parity with the broader AI ecosystem.