Gemini SEO: Citations, Sources & Optimisation

How Gemini picks its sources

Gemini uses Google's classical web index (crawled by Googlebot, the same crawler that powers classical Google Search) plus Google-Extended-gated training data. The crawler user-agents you should expect to see in your access logs are Googlebot, Google-Extended. Allow them in robots.txt if you want your content available; deny them if you do not want your content used as training data or as live citations: the distinction matters and the controls are different per engine, so read each engine's documentation rather than assuming a single flag covers everything.

Beyond the crawl, the answer-shaping layer rewards a small set of attributes that show up consistently across our own measurements and across published research: a clear declarative quick answer in the first 200 words, a single H1, descriptive H2/H3 headings that mirror likely user questions, FAQPage and Article schema where applicable, byline + last-updated date, and an outbound link mesh that signals you sit inside a knowledgeable cluster.

Practical Gemini SEO checklist

1. Quick answer block. The first 200 words should state the answer to the page's headline question in plain declarative English. Gemini preferentially cites this passage.
2. One H1, descriptive H2/H3. Heading text is a primary chunking signal for the answer-shaping layer.
3. FAQPage + Article schema. Both must be valid JSON-LD; the FAQ name field should be a verbatim user question, not a marketing reframe.
4. Byline + dateModified. Gemini down-weights undated and unauthored pages. Both should be visible to humans and present in the schema.
5. Allow the right crawler for the right surface. If your robots.txt blocks Googlebot the page cannot be retrieved for live citations: this is the user-agent that gates appearance in Gemini's answer surface. Blocking Google-Extended only opts your content out of being used as training data and does NOT remove you from live citations. Verify both in production.
6. Internal-link mesh. Link to glossary entries for every named entity on the page; link to the pillar hub at /guides/ai-seo/ for definitional context.
7. Measure citation, not rank. Track which of your URLs Gemini cites for your seed prompts on a 7-14 day cadence. Position-style metrics from classical SEO do not transfer.

What does NOT move Gemini citations

Spending on backlink networks, doorway pages, exact-match keyword stuffing, and AI-generated boilerplate are all anti-signals in Gemini's answer-shaping layer for the same reason they are anti-signals in classical Google ranking: they correlate with low-trust clusters, which the answer layer explicitly avoids citing. The honest answer is that Gemini citation correlates strongly with the attributes that already make a page good for a human reader.

Related concepts

Related queries our readers run alongside “Gemini SEO”: Google-Extended, AI Overviews, SGE, Bard SEO (legacy), Gemini app SEO. Each is covered in the glossary or in the AI SEO pillar.

For broader context on how the answer-engine surface differs from the classical SERP: including the crawler matrix across all five engines we cover: see the pillar at /guides/ai-seo/.