How do you get cited by Gemini and Google AI Mode?

Distinguish three products before choosing what to optimize. The Gemini app is a standalone assistant that can show sources and related links, though not in every answer. AI Mode is part of Google Search: it can split a question into subtopics and search for each in parallel, a method called query fan-out. AI Overviews are summaries in search results. Our AI Overviews guide covers them in detail.

These services can use information from Google Search, but they do not share every rule. Google explicitly requires indexing and snippet eligibility for AI Mode and AI Overviews. Its documentation on Grounding with Google Search in the Gemini API describes a developer interface; it does not establish every step followed by the Gemini app.

Google Search Central’s 2026 guide is titled “Optimizing your website for generative AI features on Google Search”. It is unusually direct about the mechanics: AI Overviews and AI Mode use retrieval-augmented generation — Google’s core ranking systems fetch pages from the index, and the model writes from them — plus query fan-out, which Google illustrates with “how to fix a lawn that’s full of weeds” fanning out into “best herbicides for lawns” and “how to prevent weeds in lawn”.

The eligibility rules are the ones Search has always had, with one addition. A page must be indexed and eligible to appear with a snippet, and the site must not have opted out of Search generative AI features in Search Console. Google adds that meeting every requirement still does not guarantee crawling, indexing or serving.

Google’s single strongest recommendation is editorial, not technical: create non-commodity content. Its own example contrasts “7 Tips for First-Time Homebuyers”, which anyone could write, with a first-hand account of why one buyer waived the inspection and saved money doing it. For a company that wants to be cited, that translates to publishing what only you know — your prices, your process, your measured results — rather than restating the category.

The same guide includes a “Mythbusting” section. Google identifies several kinds of work as unnecessary for its generative features, and some as potentially counterproductive.

Compare a proposal with these recommendations. If an agency sells llms.txt, content chunking or an “AI rewrite” as a requirement for visibility in Google Search, ask for the evidence. That work may serve another purpose, such as clearer navigation or documentation, but the proposal should explain that purpose.

Gemini Apps may show sources and related links within or below an answer; Google says not all answers include them. AI Mode belongs to Search and presents links with its responses. Do not assume that one service produces only mentions and the other produces visits: measure both outcomes in your own data.

Google-Extended is a separate robots.txt token governing use of content for Gemini training and certain source-grounded responses in Gemini Apps and Vertex AI. It does not control inclusion or ranking in Google Search. AI Mode and AI Overviews depend on Googlebot, indexing and snippet controls, and the Search generative AI setting in Search Console. That setting allows a site to opt out of the listed Search AI features while remaining eligible for conventional Search.

Ahrefs compared AI Mode and AI Overviews in US data from September 2025. URL overlap was 13.7% across 540,000 query pairs; a separate content analysis of 730,000 response pairs found average semantic similarity of 86%, measured with cosine similarity. Similar meaning does not imply identical sources. This was a comparison of single generations, which is another reason to repeat measurements.

Search Console’s Generative AI performance report shows how often links to your pages appeared in AI Overviews and AI Mode, broken down by page, country, device and date. It does not show clicks, queries, positions or the passage used. It combines both features, providing an official measure of appearances but no separate result for each.

The report does not cover the Gemini app or reveal the prompts behind its impressions. We therefore repeat a fixed set of 30–50 buyer questions per language in clean sessions, logging dates, responses and sources. Citation rate is the share of questions producing a link to your site; share of voice is your share of all brand mentions. Keep the counting rule for repeats consistent. In a study dated 11 November 2025, Ahrefs followed over 43,000 keywords for a month, with at least 16 AI Overviews per keyword. Between consecutive observations, 45.5% of cited sources were new on average.

Our small check on 14 September 2026 used five GEO buyer questions, each asked twice in the Gemini app and twice in AI Mode: 20 answers in total. Among the complete source URLs we could recover, one appeared in both services. For three questions, complete URLs could not be extracted from Gemini. This is a one-day observation with incomplete source data, not a representative comparison or a measure of how often either service cites a particular brand.

Ask these three questions to establish how the proposed work follows Google’s guidance.