# How Gemini, AI Mode and AI Overviews choose sources

> How Google grounds AI answers: the Search index, RAG, query fan-out, different models per surface, the October 2026 link update, and what to do.

Source: https://mentionx.ai/learn/gemini/gemini-how-google-grounds-answers  
Author: Dipuck Jones, founder of MentionX  
Updated: 5 November 2026  
Level: Practitioner

**Gemini in the app, AI Mode and AI Overviews all build answers by retrieving pages from Google's Search index and grounding the response in them. Google calls this retrieval-augmented generation (RAG), and adds query fan-out: the model runs several related searches across subtopics to gather more sources. Google says the features are rooted in its core Search ranking and quality systems, so there's no separate "AI index" to get into. But AI Mode and AI Overviews may use different models and techniques, and the Gemini app is a separate product, so each surface can cite different pages. This lesson explains the mechanics and what they mean for your content.**

> **Key takeaways**
>
> - To be eligible as a supporting link, a page must be **indexed and eligible to show with a snippet**, and your site must be **included** in Search Console's Search generative AI control.
> - Query fan-out means you can be cited for sub-questions your page answers, not only the main query.
> - Google says there are no extra technical requirements and no special schema, files or markup.
> - The 15 October 2026 update made inline links appear higher, and began testing a dedicated section for official task pages.
> - Gemini, AI Mode and AI Overviews can differ, so track each separately.

*Checked 5 November 2026 (CET) against Gemini Apps Help, the Gemini release notes, Google Workspace Help, Google Search Central, Search Console Help, the Gemini API docs and the Google blog posts cited in Sources. Google renames and moves Gemini features often, so confirm labels and limits in your own account. "Our view" marks MentionX opinion. Ledgerline is a fictional company used for illustration.*

## How does grounding work on Google's AI surfaces?

From Google Search Central's AI guidance:

| Step | What Google says happens | What it means for you |
|---|---|---|
| 1. Retrieval (RAG) | Core Search ranking systems retrieve relevant, up-to-date pages from the Search index | Your page must be indexed and rank for something related |
| 2. Query fan-out | The model issues "a set of concurrent, related queries" to fetch more results | Sub-topic coverage matters, not just head terms |
| 3. Review | Systems review specific information on the retrieved pages | Clear, specific statements are easier to use |
| 4. Response with links | Prominent, clickable links support the information | Your page is a supporting link, often inline |

Google's fan-out example: for "how to fix a lawn that's full of weeds", fan-out queries might include "best herbicides for lawns", "remove weeds without chemicals" and "how to prevent weeds in lawn".

## How do the three surfaces differ?

| | AI Overviews | AI Mode | Gemini app |
|---|---|---|---|
| Where | Top of Google results, only when "additive" | Dedicated AI tab or mode in Search | gemini.google.com and apps |
| Trigger | Often doesn't trigger | User chooses it | Every prompt; grounding when needed |
| Best for | Gist of a question | Exploration, reasoning, comparisons | Tasks, research, creation |
| Models | May differ from AI Mode | May differ from AI Overviews | Gemini app models by plan |
| Site owner control | Search generative AI control, snippet controls | Same | Google-Extended for Gemini Apps grounding |
| Measured in | Search Console (Web), Generative AI report | Same | GA4 AI Assistant channel (web), not Search Console |

See [AI Mode vs Gemini](https://mentionx.ai/learn/engines/google-ai-mode/ai-mode-vs-gemini) for the user experience differences.

## What changed on 15 October 2026?

Google's "More ways to connect with the web from Search" post announced:

1. **More inline links, higher up.** Links embedded in words or phrases in AI responses now appear higher more often.
2. **Official pages in a dedicated section (test).** For tasks with only one or two official places to do them, Google is testing a section at the top that explains why each link is useful.
3. **News carousels** for timely articles are expanding to more locations and languages, and highlight users' Preferred Sources.
4. **Richer related links and clickable creator or publisher names** that lead to profiles.

Google also said Preferred Sources usage has grown by more than 7x since the start of 2026.

*Our view:* the official-pages test matters for brands. If the task is "cancel", "log in", "pricing", "integrations" or "status", make sure your official page for it exists, is indexed, and says clearly what it's for in the title and first lines.

## What does this mean for your content?

| Mechanic | Do this |
|---|---|
| RAG uses core ranking | Keep technical SEO and indexing healthy; check URL Inspection on key pages |
| Fan-out | Cover the sub-questions buyers ask (pricing, integrations, alternatives, implementation) on relevant pages, without creating thin pages for every variation |
| Review of specific information | State facts plainly: "Ledgerline integrates with NetSuite, Sage Intacct and Microsoft Dynamics 365." |
| Inline links | Make each page the best source for a specific claim |
| Official-page section | Keep task pages clear, current and crawlable |
| Snippet eligibility | Don't put nosnippet on pages you want cited |

> **Pro move:** take your top buyer prompt and write down 5-8 fan-out queries the model is likely to run, such as "[category] pricing", "[category] for [industry]", "[you] vs [Competitor A]", "[category] implementation time" and "[category] integrations [ERP]". Google each one in normal Search and note whether you rank on page one. AI visibility for the main prompt is usually limited by your weakest fan-out ranking, and normal rank tracking is the fastest diagnostic.

## What does Google say you don't need?

From the generative AI optimisation guide, Google says you can ignore these for Google Search:

- **llms.txt and other "special" files or markup.** Google Search doesn't use them.
- **"Chunking" content.** There's no requirement to break content into tiny pieces.
- **Rewriting just for AI.** AI systems understand synonyms and meaning.
- **Inauthentic "mentions".** Spam systems target them.
- **Overfocusing on structured data.** It's not required, but still useful for rich results.

Google also warns that creating separate pages for every query or fan-out variation, mainly to manipulate AI responses, breaches the scaled content abuse policy.

## How is the Gemini app different?

The Gemini app grounds answers with Google Search when a prompt needs current information, and shows sources. Two differences matter:

- **Controls:** Google-Extended governs grounding in Gemini Apps, but it doesn't affect Search or AI features in Search ([lesson 14](https://mentionx.ai/learn/gemini/gemini-google-extended-googlebot-controls)).
- **Personalisation:** when users connect apps such as Gmail or Photos, or have chat history and memory switched on, answers can use their own context. Two people can get different brand recommendations for the same prompt.

## Worked example: Ledgerline

*Ledgerline is a fictional accounts-payable software company. Details are illustrative.*

Ledgerline appeared in AI Mode for "best AP automation software" but not for "best AP automation for NetSuite users".

| Likely fan-out query | Ledgerline rank (illustrative) | Page cited instead |
|---|---|---|
| AP automation NetSuite integration | Not top 10 | Competitor A integration page |
| NetSuite AP automation pricing | Not top 10 | Review site comparison |
| Ledgerline vs Competitor A | 2 | Ledgerline comparison page |
| AP automation implementation time NetSuite | Not top 10 | Competitor C guide |

### Walkthrough

1. The SEO lead listed six likely fan-out queries for the NetSuite prompt.
2. Normal Search showed Ledgerline had no indexed NetSuite integration page, only a logo on the integrations grid.
3. The team wrote a full NetSuite integration page with setup steps, sync fields, implementation time and a customer quote.
4. They requested indexing in Search Console and checked it with URL Inspection.
5. Six weeks later the page ranked in the top 5 for two fan-out queries and appeared as an inline link in AI Mode for the NetSuite prompt. The team logged it as correlation, not proof.

## Common mistakes

- **Looking for an AI-only fix** when the page isn't indexed or doesn't rank.
- **Making thin pages for each fan-out variation.**
- **Assuming one surface reflects all three.**
- **Using nosnippet on pages you want cited.**
- **Ignoring official task pages.**

## Try it yourself

Pick one buyer prompt. Run it in AI Mode and note the cited domains. Write down six likely fan-out queries and check your normal Google rank for each. The fan-out query where you're weakest is your first content task.

## Quick quiz

**1. What two eligibility conditions does Google state for supporting links?**

*Answer:* Indexed and eligible to show with a snippet, plus inclusion under the Search generative AI control.

**2. What is query fan-out?**

*Answer:* The model runs several related searches across subtopics to build an answer.

**3. Do you need special schema for AI Mode?**

*Answer:* No.

## What should you check before the next lesson?

- Your key pages are indexed and snippet-eligible.
- You have fan-out query lists for your top five prompts.
- Your official task pages (pricing, integrations, login, status) are clear and indexed.

Previous: [Ask Advisor: Gemini in Google Ads, Analytics and Merchant Center](https://mentionx.ai/learn/gemini/gemini-ask-advisor-ads-analytics). Next: [Google-Extended, Googlebot and the Search generative AI control](https://mentionx.ai/learn/gemini/gemini-google-extended-googlebot-controls). Or return to the [Gemini for Marketers track](https://mentionx.ai/learn/gemini/gemini-for-marketers). Related: [How Gemini picks sources](https://mentionx.ai/learn/engines/gemini/picks-sources-citations), [How AI Mode picks sources](https://mentionx.ai/learn/engines/google-ai-mode/picks-sources-citations), [How Perplexity chooses sources](https://mentionx.ai/learn/perplexity/perplexity-how-sources-are-chosen).

## Common questions

**Do I need special optimisation to appear in AI Overviews or AI Mode?** Google says no. A page must be indexed and eligible to show with a snippet, and your site must be included under the Search generative AI control. Normal SEO best practices apply.

**What is query fan-out?** Query fan-out is when the model issues several related searches across subtopics to build one answer. It is why pages can be cited for questions they do not exactly match.

**Why do Gemini, AI Mode and AI Overviews show different sources?** Google says AI Mode and AI Overviews may use different models and techniques, so responses and links vary. The Gemini app is a separate product with its own grounding.
## Sources

- Google Search Central, [AI features and your website](https://developers.google.com/search/docs/appearance/ai-features)
- Google Search Central, [Optimizing your website for generative AI features](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide)
- Google Blog, [More ways to connect with the web from Search (15 Oct 2026)](https://blog.google/products-and-platforms/products/search/updates-ai-mode-overviews-explore-web/)
- Google Search Central, [Preferred sources](https://developers.google.com/search/docs/appearance/preferred-sources)
- Search Console Help, [Search generative AI control](https://support.google.com/webmasters/answer/16908024?hl=en)
- All accessed 5 Nov 2026.

## Frequently asked questions

### Do I need special optimisation to appear in AI Overviews or AI Mode?

Google says no. A page must be indexed and eligible to show with a snippet, and your site must be included under the Search generative AI control. Normal SEO best practices apply.

### What is query fan-out?

Query fan-out is when the model issues several related searches across subtopics to build one answer. It is why pages can be cited for questions they do not exactly match.

### Why do Gemini, AI Mode and AI Overviews show different sources?

Google says AI Mode and AI Overviews may use different models and techniques, so responses and links vary. The Gemini app is a separate product with its own grounding.

---

Written and reviewed by Dipuck Jones for MentionX Learn. The HTML version of this page: https://mentionx.ai/learn/gemini/gemini-how-google-grounds-answers