AI Citations: Brand or Press? What 4 Contradicting Studies Actually Reveal (2026)

Two recent studies flatly contradict each other. One says 86% of AI citations come from brand-owned sites. The other says 84% come from press coverage and third-party sites. Both analyzed millions of ChatGPT, Gemini, or Claude answers. Neither is wrong. They just measured different things, on different samples. Once you see why, you get a more useful takeaway than a number to memorize: the right strategy depends on your industry, not on a universal rule.

What do these four studies actually show, at a glance?

Four studies published between October 2025 and July 2026 measured where generative AI's citations come from. Their numbers look contradictory: anywhere from 16% to 86% of citations classified as "brand," depending on the study and the sector. In reality, each one lights up a different part of the picture, with its own definition of what counts as a "brand" source.

The first two rows of that table look like they cancel each other out. The last two explain why they don't.

Why does Yext's study find 86% "brand" citations?

Yext analyzed 6.8 million citations across four very local industries: retail, banking, healthcare, and food service. In those categories, a large share of citations come from business listings (address, hours, reviews), which Yext counts as "brand-managed." For a local shop or a medical practice, that means your Google Business Profile carries enormous weight.

Yext breaks its 86% down into two buckets: 44% websites, 42% listings (directories, booking platforms). Reviews and social content add another 8%, forums just 2%. Each engine has its own preference: Gemini leans toward websites (52%), ChatGPT toward listings (49%), Perplexity spreads across platforms like MapQuest and TripAdvisor.

The catch is the sample. A business listing isn't a deep article explaining your expertise. It's practical data (where, when, how much). For a neighborhood business, that's decisive. For a B2B company selling software or consulting, that 86% barely means anything — there's no meaningful "listing" for that kind of business.

Why does Muck Rack's study find 84% "earned" citations?

Muck Rack analyzed 25 million links cited by ChatGPT, Claude, and Gemini across 17 industries, between July 2025 and May 2026. It calls "earned" any source a brand doesn't directly control: news articles, academic research, government sites, encyclopedic entries, and even brand content picked up and republished by a third party. That bucket is wide — wider than what most people mean when they say "press."

The May 2026 report notes that journalism alone accounts for 27% of citations, while paid or sponsored content makes up only 0.3%. In other words, when an AI cites a source, it's almost never a paid placement or disguised advertising. Muck Rack sums up the stakes for comms teams: if your brand isn't showing up in the media coverage AI is reading, it won't show up in the answers AI is giving either.

The gap with Yext comes down almost entirely to definitions. What Yext calls "listings" doesn't exist in Muck Rack's bucket. And what Muck Rack files under "earned" (government sites, academic research) has no real equivalent in the four hyper-local industries Yext studied. Both studies are right — on their own turf.

And the biggest study, covering 11.8 billion citations — what does it say?

Profound analyzed 11.8 billion citations across 8 engines (ChatGPT, Claude, Gemini, Perplexity, Grok, Copilot, and both of Google's AI modes), between April and July 2026, spanning dozens of industries. Profound's own conclusion is blunt: the gap between industries is so wide that the global average tells you almost nothing.

On average, brand sites account for 57% of citations. But that number hides swings as wide as 1-to-4.6 depending on the industry:

The same study shows the engines themselves don't behave alike. ChatGPT leans on sources a brand doesn't control 1.8 times more than Google's AI Overviews (30% vs. 17%). Conversely, Google leans on social and user-generated content 4 times more than Microsoft's Copilot. Two brands in the same industry can end up with very different results depending on which engines they're being tracked on.

The one thing Ahrefs' study of 75,000 brands settles

Only one of the four studies measured backlinks head-on, and its verdict is blunt: classic inbound links matter less and less for AI visibility. Ahrefs' study, covering 75,000 brands and millions of AI responses, measured what best predicts whether a brand shows up in ChatGPT or Google's AI modes.

The strongest signal, by far, is YouTube mentions: a 0.74 correlation with AI visibility. Next come brand mentions across the web, even without a clickable link (0.66 to 0.71), then anchor text that names your brand (0.51 to 0.63). Classic backlinks trail far behind — well below mentions.

That finding quietly echoes what Yext, Muck Rack, and Profound each show in their own way: what matters is that people talk about you, regardless of the format. A business listing (Yext), a press article (Muck Rack), a YouTube video (Ahrefs) — three very different formats, one shared trait. A third-party source, whatever form it takes, outweighs a technical link dropped into an article purely for SEO.

What should you actually take away from these contradicting numbers?

Three concrete lessons, in order of impact.

1. Look at your own industry, not the average. If you sell cybersecurity software, your brand site matters enormously (74% in Profound's study): invest in clear, sourced, easy-to-cite product pages. If you're in pharma or the public sector, press and third-party sites carry far more weight — in government and nonprofit, brand sites account for just 16% of citations: prioritize PR, industry research, and interviews.
2. Don't neglect either lever. Even in brand-dominant industries, press and third-party sites remain the second-biggest source. And even in press-dominant industries, your own site still captures the most direct purchase-intent searches — as shown in the Toronto study on AI citations. Both levers need to be worked in parallel, not one after the other.
3. Chase mentions more than links. A mention in a YouTube video, an article, a comparison guide, or a study counts for more than a technical link. Put your PR and content budget toward stories that make other sites want to talk about you, not link-building campaigns.

"No channel makes an AI cite you by magic. What counts is the real weight your brand carries in your industry's conversation — your own site and what others say about you. Work one without the other and you're playing on half the field," says Nicolas Meridjen, founder of LightSpot.

Read side by side, the four studies point the same way: what makes a source usable is its credibility, not its type. A recognized signal of expertise and trustworthiness, whether it comes from your own site or from a third party talking about you, remains the best passport to an AI citation.

What makes a page citable, regardless of industry?

Behind these industry gaps, the mechanism is always the same. An AI doesn't cite a site because it belongs to a famous brand or a major newspaper. It cites a page because it can pull a clear, verifiable, up-to-date answer out of it — whether that page sits on your own site or on a third party's site that happens to mention you.

Whether it's a cybersecurity brand's site or a healthcare news article, the most-cited pages look alike. First, an answer that stands on its own, without requiring the reader to work through the whole surrounding article: a passage that directly answers a question, with the numbers and context inside that same paragraph. Second, facts paired with their source: an unattributed number is less likely to get picked up than one credited to a named study or organization. Third, a visible publish or update date: AI engines look for the freshest information available, and a page with no visible date gives up an edge it could otherwise have had.

In practice, it starts on your own site: it's not about volume, it's about citability. Five pages that clearly answer five specific questions beat twenty pages that circle a topic without ever committing to an answer. And when you land an external mention (an article, a study, an interview), the same principle applies to whoever is writing about you: the clearer, better-sourced, and better-dated your message is when you hand it to a journalist or a partner, the more likely it gets picked up as-is — and then cited by an AI reading that article.

How do you find out where your own industry stands?

The averages in these studies give you a frame, not an answer for your brand. The only way to know whether your industry looks more like cybersecurity (brand-dominant) or pharma (press-dominant) is to look at who actually gets cited today when you ask ChatGPT, Gemini, or Perplexity your target questions.

That's exactly what LightSpot's AI citation tracking does: it runs your key queries against the engines, records who shows up in the answers — your brand, your competitors, the press, or other third-party sites — and tracks that mix over time. From there, you know whether to prioritize improving your own pages or landing mentions elsewhere. And to check whether your site is ready to be cited, whatever your industry, LightSpot's free audit runs 46 checks across 3 pages in a few minutes, no account required.