What Sources Do AI Engines Cite? Data by Topic

AI engines don't cite the same sources. What sources do AI engines cite comes down to two things: which engine answers, and what topic the question falls under. Our 500-prompt study across ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews found that citation rates varied by more than a factor of two between engines, from ChatGPT's 35.5% up to Perplexity's 100%, and that no single domain landed in every engine's top 10. What that study didn't break down is how citations shifted by topic. That's what this piece covers.

How we ran this test

We built this test at Yogoo AI to see what actually gets cited, not what marketers assume gets cited. ChatGPT, Claude, and Gemini were tested through their APIs with live web search turned on: GPT-4o-mini + web search (API), Claude Haiku 4.5 + web search, and Gemini 2.5 Flash + web search. That's not the same as the consumer apps, so treat those three numbers as a floor, not a mirror. Google AI Overviews and Perplexity are the real, live products, which means those numbers reflect exactly what a person sees when they search.

How does citation behavior differ by topic?

The 500 prompts split into 12 topic clusters, and the sources each engine reached for shifted from one cluster to the next.

"AEO Fundamentals" and "AEO Tools & Platforms" prompts leaned hardest on hubspot.com, with tryprofound.com showing up frequently in both. "AI SEO" and "AI Search Optimization" prompts pulled onelittleweb.com most often, with youtube.com and tryprofound.com also appearing near the top. The two practical how-to clusters, "How to Track Brand Mentions" and "Practical AI Visibility How-To," both put reddit.com at the top, with youtube.com in the top three of each. "Choosing a Citation Tracking Platform" pulled from youtube.com, llmrefs.com, and reddit.com.

Pricing prompts stood apart. The "Pricing & Cost" cluster cited siftly.ai, slatehq.com, and therankmasters.com most, all AI-visibility vendor domains, where most other clusters leaned toward UGC and review sites like Reddit and YouTube.

What this means for your content

Pricing content should live on a page, not a roundup.

The Pricing & Cost cluster cited vendor domains (siftly.ai, slatehq.com, therankmasters.com) directly, not third-party comparison or roundup sites. If pricing is a topic you want cited on, a clear, crawlable pricing page is more likely to get pulled than a blog post explaining pricing in general terms.

How-to content is competing with Reddit and YouTube, not other blogs.

Both how-to clusters, "How to Track Brand Mentions" and "Practical AI Visibility How-To," cited reddit.com first, with youtube.com close behind in the top three. A well-written blog post answering the same question is competing against a forum thread and a video, not against another company's how-to page. Worth treating video and community presence as part of the how-to content plan, not just written content.

Fundamentals and tool-comparison content rewards long-form authority.

"AEO Fundamentals" and "AEO Tools & Platforms" leaned on hubspot.com and tryprofound.com, both long-standing, high-authority blog domains. This is the one cluster in the data where an established, structured blog post is the format actually getting cited, which makes it the safer bet for evergreen educational content compared to how-to or pricing topics.

"Choosing a platform" content sits between the two patterns.

The comparison cluster pulled from youtube.com, llmrefs.com, and reddit.com, a mix of UGC and a niche review/directory site rather than vendor blogs. Comparison and buying-guide content may need a directory or review presence alongside a blog post to actually get cited here.

If you're trying to figure out how to monitor brand mentions across ChatGPT and Perplexity specifically, that's a narrower problem than "getting cited" in general, and it's worth treating separately.

Engine picks which pool of sources is even in play. Topic decides which one gets pulled from it. If you're optimizing for citations, what the question is about matters as much as which AI answers it.

If you're comparing platforms for tracking your own citation performance, we broke down how the major AI visibility monitoring platforms stack up against each other. And if you want to see exactly when Google AI Overviews cites your brand, that's worth checking on its own.