Who Do ChatGPT, Gemini, and Perplexity Actually Cite?

Hi everyone, this is Neo.

Lately I’ve been tracking a trend: more and more overseas buyers skip traditional Google search entirely and go straight to ChatGPT, Gemini, or Perplexity. For those of us running independent sites, that means the traffic entry points are fundamentally shifting.

But here’s the catch — if your site never gets cited by these AI engines, you’re effectively invisible on this new battlefield.

Today’s article breaks down a fresh research report from BrightEdge on the “personalities” of five AI search engines (ChatGPT, Google AI Overviews, Google AI Mode, Google Gemini, and Perplexity) when it comes to citing sources — and how we as independent site sellers should position ourselves accordingly.

1. Five AI engines, shockingly little overlap in cited sources

BrightEdge’s study covered the five most popular AI search entry points. When the researchers compared how much these engines overlap in the sites they cite, they found something fascinating:

Between any two AI engines, source overlap ranged from a low of 16% to a high of just 59%.

What does that mean? Being visible in Google AI Overviews doesn’t mean ChatGPT can see you. Getting cited by Perplexity doesn’t mean Gemini even knows you exist.

Every AI engine has its own “information diet.” They pick content from different sites to answer users’ questions.

Neo’s take: SEO used to be about watching one algorithm — Google’s. Now AI search is fragmented: it’s like having to satisfy five different “editorial tastes” at once.

2. But brand citations are remarkably consistent — that’s the real insight

The source overlap was tiny, yet the researchers found a counterintuitive result at the brand level:

Brand name overlap was much higher — at least 36%, up to 55%.

In other words, these AI engines may pull content from different websites, but the brands they end up recommending are largely the same crowd.

Why? When a user asks “what’s the best outdoor tent brand,” ChatGPT and Gemini might grab content from different review sites — but the names that keep showing up across those review sites are almost always brands with strong established recognition: Patagonia, REI, Coleman.

Neo’s take: This validates an old truth — SEO ends at branding. When users already equate a product category with your brand in their heads, AI engines can hardly avoid mentioning you. That’s the power of “branded search.” Google has actually been strengthening brand signals through systems like Navboost since 2004 — most SEOs just never noticed.

3. Three site categories, different levels of “picky” per platform

BrightEdge sorted all cited sites into three categories, and the differences are striking:

1. Institutional sites (government, academic, industry leaders)

Citation share ranges from 10% to 26%. Gemini loves these sites the most (26%); AI Overviews likes them least (10%).

2. Commercial/editorial sites (media, reviews, product listings)

This is the “staple diet” of every AI engine, ranging from 37% (Gemini) to 51% (AI Overviews). Review sites, comparison posts, and industry media coverage are the content AI grabs most often.

3. UGC sites (forums, video platforms, social content)

The most divergent category. Gemini barely trusts UGC at all (0.2%), while AI Overviews trusts it up to 18%.

Neo’s take: If you run a B2C independent site, product reviews and real customer feedback carry serious weight in AI Overviews. But if you want visibility in Gemini, Reddit threads alone won’t cut it — you need authoritative media coverage and industry report citations.

4. Personality profiles of the five AI engines

Gemini — the most conservative “academic”

  • Institutional sites: 26%
  • .gov sites: 13%
  • .org sites: 23%
  • UGC: 0.2%

Gemini’s citation strategy is very cautious, with a heavy preference for official, authoritative, nonprofit content. It would rather cite Texas Instruments’ official documentation than an electronics hobbyist forum’s real-world measurements of an op-amp chip.

The funny thing is, in practice, that forum thread’s hands-on data and comparisons might actually be more useful for a buying decision.

AI Overviews — the most “grounded”

  • UGC: 18%
  • Single video platform share: 10.6%
  • Forum platforms: 2.9%

AI Overviews trusts user-generated content — YouTube videos, forum discussions — far more than any other platform. Google’s underlying stack may be using speed-focused ranking tech like FastSearch, which makes UGC easier to crawl.

Perplexity — the most “scholarly”

  • Medical, government, encyclopedia, and publishing sources combined: 30%
  • .edu citation rate: 3.2% (the highest of all platforms)
  • UGC: 1.5%

Perplexity’s citations are heavily concentrated on authoritative sources. If your independent site is in a vertical like health or education, getting cited by .edu domains or established publishers gives you a serious edge in Perplexity.

ChatGPT — the most “open-minded”

  • Its top 10 sources account for just 18.5% of all citations

ChatGPT’s citations are the most dispersed — it doesn’t fixate on a handful of big sites. That means smaller sites genuinely have a shot at visibility in ChatGPT.

Google AI Mode — a lot like ChatGPT

  • Top 10 sources: 19.4%
  • Institutional sites: 14%
  • UGC: 7%

Google AI Mode’s citation strategy is balanced, with no extreme preferences.

Neo’s take: Different AI engines correspond to different user contexts and trust logics. Gemini rewards playing the “authority card,” AI Overviews rewards building “social proof content,” Perplexity rewards accumulating “academic and professional endorsements,” and ChatGPT rewards casting a wide net across many placements.

5. .edu sites aren’t authoritative anymore? A finding that flips conventional wisdom

The SEO world has long held a superstition: .edu and .gov domains carry inherently higher authority.

But BrightEdge’s data shows all AI engines cite .edu sites at pitifully low rates — Perplexity, the highest, is still only 3.2%.

What does that tell us?

Users don’t ask AI academic-paper-style questions. They ask “what should I buy,” “how do I use this,” “which one is better value.” University sites can’t answer those questions — review sites, forums, and video tutorials can.

Neo’s take: Stop chasing .edu backlinks out of blind habit. In the AI search era, authority has been redefined: whoever helps users solve real problems in real use cases is the authority.

6. Google’s three AI systems aren’t even one family

A lot of people assume Google AI Overviews, Google AI Mode, and Gemini are the same system in different skins. The data says otherwise:

  • AI Overviews and AI Mode source overlap: 59%
  • Gemini and AI Overviews overlap: 34%
  • Gemini and AI Mode overlap: 27%

Gemini differs more from the other two Google AI products than it does from ChatGPT. Google is running at least two completely different AI search architectures internally.

Neo’s take: Don’t optimize for “Google” as one monolithic thing. Performing well in Google AI Overviews doesn’t mean you’ll do well in Gemini. Their underlying source logic is completely different.

7. Practical strategy for independent site sellers

Based on the data, here are a few suggestions you can act on immediately:

1. Brand building is priority number one

Instead of pouring effort into gaming one AI engine’s algorithm, get users to strongly associate your brand with a category. No matter how AI engines evolve, if “best XX brand” questions make users think of you, you have the foundation for being cited.

2. Structure content for “one fish, multiple dishes”

  • Authority endorsement content (industry reports, whitepapers, official partnerships) → targets Gemini and Perplexity
  • Reviews and comparison tests → targets AI Overviews and ChatGPT
  • UGC word-of-mouth content (YouTube reviews, forum discussions, Reddit threads) → targets AI Overviews

The ROI of .edu backlinks is declining in the AI search era. Move more budget toward industry media, review sites, and PR coverage. To quote BrightEdge directly: “Investment in PR, trade coverage, review site visibility, and category comparison content translates into visibility across every engine, not just one.”

4. Watch the compliant sponsored content opportunity

BrightEdge didn’t mention this, but the original article’s author did: many authoritative sites publish clearly-labeled sponsored articles (compliant with FTC and Google guidelines). Through these compliant sponsored placements, you can strongly associate your brand and products with authoritative sites — raising your odds of being cited by AI.

Neo’s take: The essence of AI search SEO has shifted from “outsmarting the algorithm with tactics” to “owning mindshare with content.” Your content needs to appear on the sites each AI engine trusts — not just play the old traditional Google ranking game.

Summary

Here are the key findings from the BrightEdge report we covered today:

  1. The five AI engines cite wildly different sources — overlap is as low as 16%
  2. But brand recommendations are highly consistent — branding is the cross-platform “hard currency”
  3. Gemini leans conservative and authoritative, AI Overviews favors UGC, Perplexity leans academic, ChatGPT is the most open
  4. The .edu authority myth is dead — practical content is the new authority in AI search
  5. Google’s three AI systems are not one entity — don’t optimize for them as a monolith
  6. PR, review sites, industry coverage, and category comparison content are investments that pay off across every AI engine

The AI search landscape is still shifting fast, but one logic never changes: brands users remember are the brands AI cites.

Hope this sparks some ideas. If you run into any questions putting it into practice, drop them in the comments.