How ChatGPT uses the search index, and why SEO remains the foundation of GEO

ChatGPT seems to conjure answers out of thin air, but behind the scenes it leans heavily on classic search infrastructure. When ChatGPT needs an up-to-date answer, it does not build its own web index: it sends a query to Bing’s search index and retrieves candidate sources there. Those sources are then summarised and cited. In other words: whether your brand shows up in an AI answer starts with a simple question, is your page indexed and findable at all?

That has direct consequences for B2B marketers. Generative engine optimization (GEO) is often presented as an entirely new discipline, separate from SEO. In practice, strong classic SEO remains the foundation: indexation, crawlability, structure and authority feed the answers AI models generate. In this article you will read how that retrieval works exactly, what the research says about the overlap between rankings and citations, and what you should do right now.

Test your AI visibility: score your page with our GEO check.

How ChatGPT retrieves sources

ChatGPT Search has no independently built index of the web. For live information it relies on Bing’s search index. Yoast puts it sharply: “While it uses various sources, Bing is currently a major driving force.” The model first decides whether a question needs current information, retrieves it in the background through Bing’s index, and then shows the answer with source attribution. Not every question triggers that search: which prompts activate ChatGPT search and actually surface your content determines whether you are a candidate at all.

The practical conclusion is blunt: if your site is not in Bing’s index, ChatGPT cannot find your content. Yoast states it explicitly: “If your site isn’t in Bing’s index, it won’t appear in ChatGPT’s results.” Retrieval runs in roughly two layers. First there is retrieval itself, pulling candidate URLs from the search index. Then comes generation, where the model reads those sources, summarises them and selects what it cites. Indexation is therefore the gateway, and the content of your page decides whether you make the selection.

What does that mean concretely for your findability?

Anyone without control over the first layer automatically falls outside the other two. That is why the fundamentals of SEO and GEO remain tightly interwoven.

Bing as the gatekeeper, not Google

A common mistake is thinking that a strong Google position automatically results in AI citations. For ChatGPT, access runs first and foremost through Bing, not Google. That makes Bing Webmaster Tools an underrated but crucial channel: it is there that you check whether your most important pages are actually indexed.

At the same time, Bing and Google largely reflect the same underlying quality signals. Both search engines reward technically healthy sites, relevant content and authority through backlinks. A page that performs well in Google is usually well indexed and findable in Bing too. The signals you build for classic SEO therefore work double: they strengthen your position in both indexes, and with it your chance of being retrieved by ChatGPT.

In practice this means three things for B2B:

• Check your indexation in Bing Webmaster Tools, not just in Google Search Console.
• Allow the relevant AI crawlers in your robots.txt, such as OAI-SearchBot, so your content can be surfaced.
• Keep building authority, because it is a signal that counts in both worlds and that helps you rank higher on Google while making you more visible in AI answers.

What the research says about rankings and citations

This is where it gets nuanced. Indexation is necessary, but a high ranking is no guarantee of a citation. Ahrefs analysed 15,000 long-tail queries across four AI assistants and found that only 12% of the URLs cited by AI also appear in the Google top 10 for the original question. The overlap with the Bing top 10 was 10%. For ChatGPT specifically the overlap was even lower: 8% for in-text citations and 6.1% for the references.

The numbers differ sharply per platform:

The explanation Ahrefs gives: AI assistants use query fan-out, they split your question into multiple query variants and retrieve sources for each of those variants instead of optimising for your exact prompt. As a result they often cite pages that rank nowhere near the top for the original search term. Google AI Overviews are the exception, at 76% overlap, because they function as an extension of the classic search results.

The lesson is not “rankings do not matter”, but “rankings are not enough”. You need to be findable across a broader set of related questions, not just on one head keyword.

Why SEO remains the foundation of GEO

If the ranking correlation is low, why does SEO remain the base? Because the SEO fundamentals deliver precisely the signals that decide whether you make the candidate list. A Semrush study of 5 million cited URLs showed that the most-cited pages use structured data strikingly often. With ChatGPT, roughly 25% of cited pages carried Organization schema and 20% Article schema; for Google AI Mode those figures were 34% and 26% respectively.

Semrush emphasises firmly that these are correlations, not direct causality. According to the researchers, technical factors work as “a foundation that enables visibility when other signals align”. That is exactly the role of SEO within GEO: it is the foundation that makes visibility possible once the content signals line up. Concretely, you build that foundation with:

• Indexation and crawlability: without an index there is no retrieval, in any search engine.
• Structured data: schema markup helps both classic search engines and AI models understand your content.
• Authority: backlinks and consensus across multiple sources make you a reliable choice to cite.
• Clear, citable content: if you want to structurally become source material for AI, your text has to be factual, chunked and easy to lift. How direct, clear phrasing increases your chance of a citation is covered in direct response copy for LLM optimisation.

GEO and SEO are therefore not an either-or choice. The difference between them is explained in GEO vs SEO, but the core stays the same: you build the technical and authority base first, and on top of that you lay the GEO layer that makes your content citable.

What to do right now

For a B2B organisation that wants to be visible in both Google and ChatGPT, the approach is pragmatic:

1. Check indexation in both worlds. Use Google Search Console and Bing Webmaster Tools to spot gaps.
2. Allow the right crawlers. Permit OAI-SearchBot and other relevant AI crawlers in your robots.txt.
3. Think in question variants, not one keyword. Cover the broader set of related questions your audience asks, given the query fan-out of AI models.
4. Add structured data. Start with Organization and Article schema on your most important pages.
5. Make your content citable. Concrete figures, clear headings and short, factual passages get picked up more easily.
6. Keep building authority. Backlinks and mentions count in both worlds.

Anyone with this base in order increases not only their Google visibility, but also the chance that ChatGPT and other AI models retrieve and cite the brand.

Frequently asked questions

Does ChatGPT use Google’s index or Bing’s?

For live searches, ChatGPT Search relies mainly on Bing’s search index, not Google’s. According to Yoast, your site will not appear in ChatGPT’s results if it is not in Bing’s index. Bing indexation is therefore the first condition for AI visibility through ChatGPT.

Does a high Google position mean ChatGPT will cite me?

Not automatically. Research by Ahrefs across 15,000 queries shows that only 12% of the URLs cited by AI appear in the Google top 10, and even fewer for ChatGPT specifically. A good ranking helps, but you also need to be findable across the broader set of related questions AI models ask through query fan-out.

Is SEO still useful for AI search engines?

Yes, and it remains the foundation. SEO delivers indexation, crawlability, structure and authority, precisely the signals that decide whether you make an AI’s candidate list. Semrush describes technical factors as the base that makes visibility possible once the other signals line up.

Which quick win delivers the most for AI visibility?

Start with indexation: check in Bing Webmaster Tools whether your most important pages are indexed and allow the relevant AI crawlers in your robots.txt. Without indexation no AI can retrieve your content, however strong that content may be. After that you add structured data and citable passages.

Need help?

Want to translate this into execution? See how we approach it with AI findability.