One Brand, Five Answers: How ChatGPT, Perplexity, Gemini, Copilot, and Google AI Mode Each Choose Sources

Article 7 of 9 in the 2027 AEO + GEO series

Run the same buyer question in five AI assistants and you will usually get five different shortlists. A brand can be the first recommendation in one, a footnote in another, and missing entirely from a third. That is not a bug in your testing. These systems retrieve from different indexes, weigh sources differently, and present citations in different ways.

This article explains the main differences, what they imply for your strategy, and how to avoid optimizing for one platform at the expense of the rest.

A note on certainty before we start: none of these companies publishes a complete account of how it selects sources, and the systems change often. What follows combines what the platforms have said publicly with patterns reported by SEO researchers and tool vendors. Treat the specifics as working hypotheses and confirm them with your own tests.

The big picture: three source pools

Most differences trace back to where each assistant gets its information when it searches the live web.

1. Google's ecosystem. Google AI Overviews, Google AI Mode, and Gemini draw on Google Search and its related systems, including the Knowledge Graph.
2. Bing's ecosystem. Microsoft Copilot is built on Bing. ChatGPT search is widely reported to rely substantially on Bing's index, supplemented by OpenAI's own crawlers and other partners. OpenAI has not published a complete breakdown, so treat that as a strong hint rather than a confirmed rule.
3. Independent indexes. Perplexity operates its own crawler and index and also uses other sources, and it is built around showing citations for nearly every answer.

Across all of them, the model's training data also shapes what it knows about your brand before any search happens. That is one reason long-term reputation work matters even for platforms that search live.

Platform by platform

ChatGPT

• How it finds sources. ChatGPT uses web search for many, though not all, queries. When it searches, it is widely reported to lean on Bing's index and OpenAI's own crawlers. For queries answered without search, it relies on what it learned in training.
• Citation behavior. Citations appear when search is used, but their presentation and consistency vary. Many brand mentions come with no link at all.
• What it tends to favor. Studies from SEO vendors report heavy use of broadly corroborated sources such as Wikipedia, Reddit, review and comparison sites, and editorial coverage. Brands mentioned across third-party sources tend to fare better.
• What to do. Confirm your site is indexed in Bing, allow OpenAI's documented crawlers where appropriate, and invest in earned coverage and listings that Bing indexes. Check your robots.txt and CDN settings so you do not block crawlers by accident.

Perplexity

• How it finds sources. Perplexity retrieves in real time on each query, using its own crawler, PerplexityBot, along with other data sources.
• Citation behavior. Numbered citations appear by default, which makes it the easiest platform to audit.
• What it tends to favor. Reported patterns include fresh content, direct answers near the top of a page, specialist sources with precise pages on the topic, and review and comparison sites for commercial queries. Reddit shows up frequently in some analyses.
• What to do. Make sure PerplexityBot can crawl priority pages, lead sections with the answer, keep content current, and maintain presence on review and community sources. Because it responds quickly to new content, it is a good place to test whether changes are being picked up.

Gemini

• How it finds sources. Gemini grounds answers in Google Search and Google's knowledge systems when it uses live information.
• What it tends to favor. Sources Google already indexes and trusts, and entities Google can resolve. One analysis reported only modest overlap between Gemini's cited sources and traditional ranked results, so strong rankings help but do not guarantee inclusion.
• What to do. Build a clear entity presence, with a consistent Organization and Person identity, a Google Business Profile where relevant, and schema that matches visible content. Keep fundamentals strong in Google Search Console. Note that Google's Google-Extended setting controls certain AI uses of your content but does not change how pages appear in Search.

Microsoft Copilot

• How it finds sources. Copilot retrieves from Bing, and Microsoft also draws on professional and enterprise context in some experiences.
• What it tends to favor. Pages that rank and are indexed in Bing. Analysts also report that LinkedIn content is cited prominently in some professional queries.
• What to do. Set up Bing Webmaster Tools, submit sitemaps, adopt IndexNow, claim Bing Places for local businesses, and keep company and author profiles current on LinkedIn. Bing's AI Performance reporting can show how often your pages are cited in Copilot.

Google AI Mode and AI Overviews

• How it finds sources. Both are built on Google's index. AI Mode in particular uses query fan-out, breaking a question into many sub-queries and retrieving passages for each, as covered in article 3.
• What it tends to favor. Strong SEO fundamentals remain the prerequisite, but because of fan-out, many cited pages are not the top-ranking result for the original query. Reported patterns include a heavy presence of YouTube, Reddit, LinkedIn, and listicle or comparison pages, along with pages that have rich structured data.
• AI Overviews versus AI Mode. Analysts report these two cite noticeably different sets of sources, so treat them as separate surfaces in your testing.
• What to do. Cover the full question family around each topic, write passage-level answers, strengthen structured data and entity signals, and use Search Console's Generative AI reporting alongside your own prompt tests.

Side-by-side summary

This table simplifies fast-moving systems. Use it to decide where to look first, not as a rulebook.

What the overlap data means

Studies that compare cited sources across platforms consistently find limited overlap. One analysis cited in industry write-ups found that only about one in nine domains was cited by both ChatGPT and Perplexity, and others report low overlap between AI Mode and AI Overviews. The exact numbers depend on the prompts, categories, and time period, so do not quote them as constants. The direction is consistent, though: appearing in one assistant is a poor predictor of appearing in another.

That has two consequences:

• You need to measure every platform your buyers use, not just the one that is easiest to check.
• Foundations that help everywhere, such as crawlability in both Google and Bing, clear entities, passage-level answers, and strong third-party presence, should come before platform-specific tweaks.

A practical approach

1. Find out where your buyers actually ask. Use customer interviews, referral data in GA4's AI Assistant channel, and surveys to decide which platforms deserve the most attention.
2. Fix cross-platform foundations first. Allow approved crawlers, get indexed in both Google and Bing, use IndexNow, and keep key content in rendered HTML.
3. Build a prompt set of 15 to 30 buyer questions covering discovery, comparison, and constraints, and run it on each platform on a schedule. Repeat each prompt several times, since answers vary.
4. Log every citation by platform. Note which domains and pages each assistant uses for your category.
5. Look for platform-specific gaps. If Perplexity cites you and ChatGPT does not, check Bing indexing and third-party coverage. If Copilot ignores you, check Bing rankings and LinkedIn. If Gemini misdescribes you, check your entity data.
6. Make targeted changes, then retest. Change one thing at a time where you can, so you know what moved results.
7. Revisit quarterly. Platforms change retrieval systems and citation behavior often, so last quarter's findings may be out of date.

Beyond these five

Other assistants matter too, depending on your audience. Claude, Meta AI, and Apple's and Amazon's assistants each have their own retrieval arrangements and sometimes work with different search or data partners. The method above works for any of them: identify the source pool, test your prompts, log citations, and fix gaps.

Common mistakes

• Optimizing only for Google. Being indexed and healthy in Bing is a separate job, and it may decide visibility in ChatGPT and Copilot.
• Trusting one answer. Responses vary run to run, so judge visibility from repeated tests.
• Treating reported platform rules as fixed. Many claims, such as which index powers which assistant, come from outside observation and can change without notice.
• Ignoring AI Overviews versus AI Mode. They are different surfaces with different citation patterns.
• Blocking crawlers unintentionally. Overly strict robots.txt, WAF, or CDN rules can cut you off from some systems while leaving others untouched.
• Chasing hacks. Tactics aimed at one platform's quirks tend to expire. Evidence, clarity, and corroboration last.

Quick FAQ

Which platform should I prioritize? The one your buyers use most. Check your analytics, ask customers, and test your prompt set in each platform before deciding.

Do I need a separate strategy for each platform? Mostly no. A shared foundation covers the majority of the work, with a short list of platform-specific actions on top.

Does ranking well on Google help in ChatGPT? It can help indirectly through brand authority and coverage, but if ChatGPT's retrieval relies on Bing, pages that are poorly indexed there may be missed. Check both.

How often do these behaviors change? Often. Re-verify against official documentation and fresh tests regularly.

Key takeaways

• Different assistants retrieve from different source pools, so visibility on one does not predict visibility on another.
• Google powers AI Overviews, AI Mode, and Gemini. Bing powers Copilot and is widely reported to play a major role for ChatGPT. Perplexity runs largely on its own retrieval.
• Fix cross-platform foundations first, then use citation logs to find platform-specific gaps.
• Treat reported behaviors as hypotheses and verify them with repeated, logged tests.
• Re-test regularly, because platforms change their systems without announcement.

Next in the series: [Freshness Signals: How Often to Update Content for Generative Search (and What Counts as an Update)](/articles/freshness-signals).

For the full picture, start with our main guide: AI Visibility for Local Service Businesses.