Quick answer: AI answer engines mostly cite independent third-party sites, not the brand itself: a 2026 Semrush analysis puts external citation share at 82.3% for SaaS, 79.4% for finance, and 69.6% for ecommerce. Finance is nearly pure-referee; consumer electronics is brand-dominated, with Apple near 60% self-citation. Which bucket your brand falls into should decide your next AI-visibility budget.
Key takeaways
- Brand-owned domains are a minority of AI citations in SaaS (17.7%), finance (20.6%), and ecommerce (30.4%): most citations go to independent third parties.
- Finance is nearly pure-referee (NerdWallet and Investopedia outcite individual banks); consumer electronics is brand-dominated (Apple ~60% self-citation).
- Some query types are structurally locked out. Legal and regulatory questions route to sources like .gov domains regardless of content quality, because of how AI query fan-out works.
- A citation and a mention are different signals. Only 28% of brands get both (a footnoted link and a name-check) in the same AI answer.
- Run your own audit. Test 15-20 prompts across ChatGPT, AI Mode, and Gemini before assuming the category average applies to your brand.
Someone on your marketing team is going to ask this quarter whether the company shows up in ChatGPT, and the honest answer is usually more complicated than yes or no. Most brands assume that if their content is good enough, AI answer engines will eventually cite their own site as the source. The data says otherwise for most categories. AI systems overwhelmingly cite independent third-party sites, not the brand itself, and the split is different enough by industry that a generic AI-visibility strategy wastes budget in the wrong place. This article breaks down the actual citation-share numbers by category, explains why some industries are brand-dominated while others are almost entirely referee-driven, and gives you a way to find out which bucket your own brand falls into before you decide whether the next dollar goes to schema and on-page work or to PR and digital citations.
What 'brand vs. referee' means in an AI answer
A citation is a footnoted source link inside an AI-generated answer; a mention is your brand named directly in the answer text, with or without a link back to your site. They are not the same signal, and most brands only track one of them, if either. Rankscience's analysis of citations versus mentions found that only 28% of brands studied achieved both in the same AI response. A brand can be named as a recommendation while the actual supporting link points somewhere else entirely.
A "referee" site is any independent third party cited instead of the brand's own domain: a review platform, an industry publication, a comparison site, a forum thread, or a government resource. When an AI system answers a category question ("best CRM for small teams," "is this bank's savings rate competitive," "which laptop should I buy"), it's choosing between citing the company that makes the thing and citing someone else's independent judgment about the thing. Which one it picks, most of the time, isn't really about your content quality. It's about your category.
The citation split by industry: SaaS, finance, and ecommerce
Brand-owned domains are a minority source of AI citations across every major commercial category measured so far. Aleyda Solis's analysis of Semrush Enterprise data (Semrush's enterprise-tier SEO and AI-visibility analytics platform) puts external (non-brand) citation share at 82.3% for SaaS, 79.4% for finance, and 69.6% for ecommerce: brand-owned sites account for only 17.7%, 20.6%, and 30.4% of citations in those categories, respectively. A separate B2B SaaS citation benchmark study tested 40 product categories and found ChatGPT cited the recommended tool's own site only 11.6% of the time, even when it was actively recommending that exact tool. Two independent data sources landing in the same range is a strong signal this isn't a fluke of one dataset.
| Industry | Brand-owned citation share | External (referee) citation share |
|---|---|---|
| SaaS | 17.7% | 82.3% |
| Finance | 20.6% | 79.4% |
| Ecommerce | 30.4% | 69.6% |
Ecommerce comes closest to balanced, but even there, roughly seven out of every ten citations go to someone other than the brand. If your AI-visibility plan assumes your own product pages and blog posts are the thing AI systems will cite, this table is probably why that plan is underperforming.
Why finance is pure-referee and consumer electronics is brand-dominated
Finance and consumer electronics sit at opposite ends of the same spectrum, and the reason why tells you something useful about your own category. A 22-category breakdown of AI citation patterns found brand-owned sources dominate in 13 categories and independent referee sites dominate in 8 (with one mixed), and the two clearest examples are finance and consumer electronics. In finance, comparison sites NerdWallet and Investopedia (independent personal-finance publishers, not banks) each account for roughly 14-15% of ChatGPT citations on finance queries, more than any individual bank's own site. Consumer electronics runs the opposite way: Apple's own domain accounts for roughly 60% of citations about Apple products.
The pattern makes sense once you see it. Finance queries carry regulatory and trust baggage, so a reader asking about savings rates or loan terms wants a neutral third party's comparison, not the bank's own marketing page, and AI systems default to the referee. Consumer electronics queries are usually about specifications, compatibility, or official documentation, where the manufacturer is the only authoritative source, so AI systems default to the brand. Your category's position on that trust-versus-authority spectrum predicts your citation split better than your content budget does.
Some queries are structurally locked out of brand citations
No amount of content quality moves certain query types toward brand-owned citation, because the AI system restricts which domain types are even eligible before it evaluates your content at all. Lily Ray's fan-out analysis found that ChatGPT's query fan-out, the process of splitting one question into several sub-queries and pulling sources for each, assigns different eligible source types by query type. In a 20-prompt legal test, nearly all citations came from .gov domains. Product and pricing queries pulled brand-owned domains. Opinion and review queries pulled Reddit.
This matters because it separates two different problems that look identical from the outside. If your brand isn't cited on a legal or regulatory question, more content won't fix that: the query is structurally routed to government sources regardless of what you publish. If your brand isn't cited on a review-style question, the fan-out is routing to community sources like Reddit, which means your fix is community presence, not a better blog post. Knowing which of these you're dealing with before you spend budget is the entire point of running the audit in the next section.
How to find your brand's own citation split
The fastest way to know your actual citation split is to run it yourself rather than assume the industry average applies to your brand. Pick 15-20 prompts that represent how a real buyer would ask about your category, spread across ChatGPT, Google AI Mode, and Gemini. Log every domain each engine cites, then tag each one as brand-owned or third-party, and within third-party, split further into publisher, review site, or community/social. Once you total it up, you have a real, brand-specific citation share percentage instead of a category average.
This is worth turning into an ongoing check, not a one-time snapshot. Bing Webmaster Tools now ships a native Citation Share metric alongside Topics, Intents, and Compare views, giving you a free way to track how your own citation percentage moves over time without re-running the manual prompt list every month. Treat the manual audit as your baseline and the ongoing metric as your trend line. If you want a faster read than building the tracker yourself, checking whether your brand is cited in ChatGPT walks through the same prompt-logging method in more detail.
Where to put your budget: schema and on-page, or PR and digital citations
Once you know your split, the budget decision becomes mechanical instead of a guess. If your category or your specific audit results skew brand-dominated, closer to the consumer-electronics pattern, the higher-leverage spend is on-page structure, schema markup, and making your own site the cleanest, most extractable source on the topic, since AI systems already default to citing brands like yours directly. If your category skews referee-dominated, closer to the finance pattern, that same spend has a much lower ceiling, because AI systems are structurally routing around your domain regardless of how well-structured it is. There, PR, digital citations, and building a real presence on the review and community sites that already dominate your category's citations will move the needle further than another round of on-page optimization.
Treat all of this as directional, not definitive. These are benchmark figures from one analyst's read of one enterprise dataset, and 15-20 prompts is a snapshot, not a census. Validate the pattern against your own audit before you move real budget, and re-check it quarterly since citation patterns shift as AI systems change how they fan out queries.
The brand-vs-referee split isn't a philosophical question about content quality. It's a measurable pattern that's different by industry, different by query type, and different for your specific brand than the category average. SaaS, finance, and ecommerce all skew heavily toward referee citations right now, finance almost entirely so, while categories like consumer electronics remain brand-dominated. Before you spend the next quarter's AI-visibility budget, run the 15-20 prompt audit and find out which side of that line your brand actually sits on.
The bottom line: If you want that audit done for you, methodically, across your real category prompts, Uygen's AI Visibility Audit finds your brand's actual citation split and turns it into a prioritized recommendation instead of a guess.
FAQ
What is the difference between a citation and a mention in AI search?
A citation is a footnoted source link inside an AI-generated answer that a reader can click through to your site. A mention is your brand named directly in the answer text, which can happen with or without a link back to you. They're tracked differently and don't always occur together. Rankscience's analysis found only 28% of brands studied got both in the same response, meaning a brand can be recommended by name while the actual cited source is a third party.
Why does ChatGPT cite other websites instead of my company's site?
In most categories, ChatGPT and other AI engines default to independent third-party sources because they're treated as more neutral than a brand's own marketing content. Semrush Enterprise data shows this is category-dependent: SaaS, finance, and ecommerce all skew heavily toward external citations, while categories like consumer electronics skew brand-dominated. Some of this is also structural: query fan-out routes certain query types, like legal and regulatory questions, almost entirely to government or authoritative third-party sources regardless of your content.
Which industries get cited from their own site the most in AI answers?
Consumer electronics is the clearest brand-dominated example, with Apple's own site accounting for roughly 60% of its own citations. A 22-category breakdown found brand-owned sources dominate in 13 categories overall, while independent referee sites dominate in 8. Categories tied to specifications, compatibility, or official documentation tend to be brand-dominated, while categories carrying regulatory or trust weight, like finance, tend to be referee-dominated.
Should I invest in PR or on-page SEO for AI visibility?
It depends on which bucket your category falls into, and you should verify this with your own data rather than assuming. If your citation audit shows brand-owned domains already dominate your category, on-page structure and schema markup are the higher-leverage spend. If referee sites dominate, PR and digital citation-building on the review and community sites that already control your category's citations will outperform additional on-page work.
Can one article change my brand's citation share?
Rarely on its own. Citation share is driven more by category-level trust patterns and query fan-out routing than by any single piece of content. A well-structured, citation-worthy article can improve your odds within the queries that are already eligible for brand-owned citation, but it won't move queries that are structurally routed to referee sources, like most legal or regulatory questions. Budget allocation based on your actual citation split matters more than any individual piece of content.
Don't know if your category is brand-dominated or referee-dominated?
The AI Visibility Audit runs the category-prompt test for you and maps your actual citation split, so budget goes to schema and on-page work or to PR and digital citations, not a guess.