Comparison matrix: how every AI search engine picks its sources in 2026

AI search engines do not choose their sources the same way. ChatGPT and Copilot pull most of their live information from the Bing index, Gemini builds on Google Search, Perplexity uses its own crawler and index, Claude relies on an external search provider, and Grok combines the open web with data from X. This article lays those differences side by side in a single matrix, then briefly explains, engine by engine, what that means for your visibility.

The goal is deliberately concise: a citable reference framework, not an optimization guide. If you want to know how to act on it, read the complete GEO guide.

The comparison matrix at a glance

The rest of the article briefly comments on each row. One warning first: these engines change fast. Treat the matrix as a snapshot of 2026, not as a law.

How does ChatGPT pick its sources?

ChatGPT pulls most of its live web information from the Bing index, thanks to the partnership between OpenAI and Microsoft. When a question calls for recent or specific information, ChatGPT breaks it down into sub-questions, retrieves candidate pages for each one from Bing, and then selects the sources it cites.

In practice, that means your findability in Bing is a direct lever for citations in ChatGPT. On top of that, the model weighs how well your page is structured and how clearly it answers a concrete sub-question. For the practical approach, see getting found in ChatGPT.

How does Google Gemini pick its sources?

Gemini grounds its answers via Grounding with Google Search: the model decides for itself whether a search is needed, runs it on Google, and processes the results with inline source annotations. The index underneath is therefore Google’s own, with all the reach that comes with it.

The consequence is logical: the signals that make you strong in classic Google results, namely relevance, quality and freshness, largely carry over into what Gemini selects. That same grounding also feeds the AI answers at the top of Google. If you want a grip on those, read getting found in Google AI Overviews.

How does Microsoft Copilot pick its sources?

Copilot works, like ChatGPT, on the Bing index, but with its own layer of internal grounding queries: the search phrases Copilot devises itself to retrieve relevant pages. Bing supplies the candidates, the model reads them and attaches citations to the sources it leans on.

What sets Copilot apart is where it lives: built into Windows and into Microsoft 365. For B2B that is relevant, because it sits close to the workplace where decision-makers already spend all day. Since 2026, Bing Webmaster Tools additionally shows how often your site is cited in those AI answers, which makes follow-up more measurable.

How does Perplexity pick its sources?

Perplexity stands out by using its own crawler and index rather than relying entirely on an external search API. That lets it include very recent pages too. Citing is in the product’s DNA: every answer contains numbered source links.

Under the hood runs a retrieval and re-ranking pipeline that scores candidate pages on direct relevance and freshness. Pages with clear, easy-to-extract passages stand a better chance of ending up as a citation. You can read the practical side in getting found in Perplexity.

How does Claude pick its sources?

Claude has been able to search the web since 2025 and adds inline citations to those answers, so you can check the source. The search layer does not run on an index of its own, but on an external search provider. According to several reports that is Brave Search; treat that detail as likely, not as officially confirmed.

Claude stands out for its caution: it leans towards verifiable claims, primary sources and balanced phrasing, and it is more reserved about citing than some other engines. For B2B content that rests on facts and nuance, that is good news, because solid, fact-checkable sources are exactly what the model likes to draw on.

How does Grok pick its sources?

Grok, from xAI, combines two sources: a broad search function on the open web and a separate search function on X. That second layer makes Grok strong on current events and trending discussions, because it pulls live signals straight from the platform. Citations are gathered automatically from the search tools and shown with the answer.

For most B2B topics, Grok is today a niche player next to ChatGPT, Gemini and Copilot. But for subjects where the public conversation moves fast, visibility on X can help determine whether you show up in Grok.

What does this mean for your B2B visibility?

The engines differ in index and emphasis, but they weigh the same things surprisingly often: a clear structure, recognizable authority, primary sources and passages that are easy to extract. That is good news. You do not need to build six separate strategies; one solid approach works across several engines at once.

Two fundamentals do most of the work. First, entity consistency: make sure your brand is recognized everywhere as the same, unambiguous entity. Second, brand mentions over backlinks: in AI search engines, where and how often you are mentioned often weighs more heavily than classic links alone.

At Customer Impact we steer on what counts for B2B: leads and revenue, not isolated vanity figures. We do not promise fixed positions (no one can do that honestly with models that choose their own sources), but we do systematically build the signals that genuinely improve your chances.

The short summary

ChatGPT and Copilot run on Bing, Gemini on Google, Perplexity on its own index, Claude on an external search provider and Grok on the open web plus X. Almost all of them cite inline, and they weigh largely the same quality signals. Whoever consistently shows up as a clear, trustworthy and extractable source wins visibility across the engines.

Want to know where your brand stands in these engines right now and where the gains are? Book your free intake or explore our AI visibility service.