How Gemini Chooses Sources and Why It Differs From AI Overviews and AI Mode

Gemini chooses sources by deciding whether a question needs a search, running its own queries against Google's search index, and citing some of the pages it drew on. If you run marketing at a SaaS or AI startup and your brand shows up in Google AI Overviews but not in the Gemini app, or the other way round, this explains why the three Google surfaces disagree and what keeps them aligned. ChatGPT, Perplexity, Claude and Copilot each have their own article. This one covers the three surfaces Google builds, Gemini, Google AI Overviews and Google AI Mode.

How Gemini chooses sources when it grounds an answer

The Gemini app grounds its answers in Google's search index. Google's documentation for its common crawlers defines grounding in Gemini Apps as providing content from the Google Search index to the model at prompt time to improve factuality and relevancy. A page Google has not indexed cannot be a Gemini source.

Google does not publish a step-by-step account for the consumer app, but it does for developers. Its Grounding with Google Search documentation, updated in September 2026, describes the flow below, which is the closest public description of what happens inside the app.

How a Gemini model grounds an answer, per Google's developer docs

The second step matters most for a brand. Gemini searches with queries it writes, not with the buyer's prompt, so the pages it reads are the ones that rank in Google for those shorter queries. That is one reason Gemini's citations rarely match the top results for the original prompt: Ahrefs found in August 2025 that just 8.6% of Gemini's citations ranked in Google's top 10 for the prompt as asked.

Why does Gemini sometimes show no sources?

Because Gemini answers some questions without searching, and not every answer carries links. Google's help article on sources in Gemini Apps says not all responses include related links or sources, and that if no Sources button appears, Gemini did not provide links for that response. When Gemini quotes a large amount of text directly from a webpage, that page appears in the sources list.

The same help article says sources can include public websites, files a user uploaded, and Google Workspace documents or emails when Workspace is connected. For a brand, that means two kinds of Gemini answer: grounded ones, which you influence through Google rankings, and ungrounded ones, which reflect what the model learned in training and which only change as the wider web describes you differently over time.

How is Gemini different from Google AI Overviews and Google AI Mode?

All three draw on Google's index, but they run different models in different places and answer different kinds of questions. Google's AI features documentation says Google AI Overviews and Google AI Mode may use a query fan-out technique, issuing multiple related searches across subtopics, and that because they may use different models and techniques, their responses and links vary. Google AI Overviews appear only when Google's systems judge them additive to classic results.

Swipe sideways to see every column

Surface	Where it appears	How it finds sources	Citations per run (Profound, May 2026)	Publisher control
Gemini app	gemini.google.com and the Gemini mobile apps	Decides whether to search, then grounds in content from Google's index	6.6	Google-Extended in robots.txt
Google AI Overviews	Above results on some Google searches	Query fan-out across Google's index; shown only when judged additive	11.1	Googlebot rules plus nosnippet, max-snippet or noindex
Google AI Mode	A separate conversational tab in Google Search	Query fan-out for exploration and comparisons	15.2	Googlebot rules plus nosnippet, max-snippet or noindex

The mechanics of the two Search surfaces are covered in how Google AI Overviews choose which sources to cite and how Google AI Mode differs from AI Overviews. The point for this article is that Gemini is the odd one out: fewer citations, a different control and, as the data shows, a different set of brands.

How often do Gemini, Google AI Overviews and Google AI Mode cite the same brands?

Not often. Each Google surface builds a noticeably different shortlist from the same index, and Gemini overlaps least with the other two.

Profound also found Gemini leaned more on editorial and reference domains, while Google AI Overviews and Google AI Mode relied more on social and user-generated platforms. Even the two Search surfaces diverge at page level: Ahrefs' December 2025 study found only 13.7% of citations overlapped between Google AI Overviews and Google AI Mode across 540,000 query pairs, although their answers had 86% semantic similarity across 730,000 response pairs. Across platforms the split is sharper still: BuzzStream's June 2026 study of 30,000 citations from ChatGPT, Gemini, Google AI Overviews and Google AI Mode found 76.1% were unique to a single platform.

Does blocking Google-Extended remove you from Gemini?

It removes your content from grounding in the Gemini app, and leaves Google Search and its AI features untouched. Google's crawler documentation says Google-Extended controls whether content Google crawls may be used to train future Gemini models and for grounding in Gemini Apps, and that it does not affect a site's inclusion in Google Search or act as a ranking signal.

Google AI Overviews and Google AI Mode follow Search controls instead. The AI features documentation says robots.txt rules for Googlebot are the control for Search, and that nosnippet, data-nosnippet, max-snippet and noindex limit what appears in AI features. So a company that blocked Google-Extended to keep content out of training has also opted out of being a Gemini source, while still appearing in Google AI Overviews. That is a legitimate choice, but it should be a deliberate one.

# Opts out of Gemini training AND Gemini app grounding
User-agent: Google-Extended
Disallow: /

# Search, Google AI Overviews and Google AI Mode follow Googlebot
User-agent: Googlebot
Allow: /

What should a SaaS brand do to be cited across all three?

Win the shared foundation once, then manage the differences. The foundation is a Google-indexed, snippet-eligible site with clear entity facts; the differences are the Google-Extended decision, the source types each surface prefers and separate tracking.

Six steps for consistent visibility across Google's AI surfaces

Google says you do not need new machine-readable files, AI text files or special schema to appear in its AI features, so skip anyone selling those as a Gemini fix. What helps a model describe a product consistently is covered in which schema markup helps AI search engines understand your product, and the cross-engine tracking setup is in how to track brand mentions across ChatGPT, Perplexity, Gemini and Google AI.

Is Gemini worth separate attention from Google SEO?

Separate tracking, yes; a separate strategy, rarely. SE Ranking's June 2026 study of 101,574 websites found Gemini was the second-largest source of AI referral traffic in 2026, at 11.56%, after growing 231% year over year. It overtook Perplexity in January 2026 in the US, the UK and globally.

Most of what earns Gemini citations is the same work that earns organic rankings on Google and Bing, with Google's index the one that counts here, plus clear entity facts, so the marginal Gemini work is small: the Google-Extended decision, editorial coverage and measurement. If Gemini already drives a meaningful share of your AI referrals, that is where a Gemini SEO agency adds value. If not, fold it into your Google work and watch the numbers.