How often does AI cite a brand's own website?

A brand-owned citation is a cited page on the same registered domain as the recommended brand's primary website. This includes a documentation or regional subdomain. Of 1,060,513 eligible links between ranked brand recommendations and cited pages, 288,629 were brand-owned. The exact rate was 27.2160%.

Semrush reports that mentioned brands and cited domains can overlap as little as 30% on Gemini. Ahrefs found that about 28% of more than 31,000 Ahrefs mentions across six engines linked to Ahrefs.com. This study counts something different, but it confirms the same boundary across many more brands: a recommendation does not mean the brand's website supplied the citation.

Third-party pages supplied 771,884 of the 1,060,513 eligible citation links, or 72.7840%. Brand-owned pages supplied the remaining 288,629, or 27.2160%.

A brand's own site is a meaningful part of the evidence mix, but it is not the whole source strategy. Marketers should maintain accurate product and documentation pages while also auditing the independent pages AI cites about the brand.

ChatGPT Search used a brand-owned page for 148,484 of 305,794 eligible citation links, or 48.5569%. Google AI Mode did so for 140,145 of 754,719 links, or 18.5692%. The difference was 29.9877 percentage points, and the ChatGPT Search rate was 2.61 times the Google AI Mode rate.

One combined owned-citation rate hides a large engine difference. Audit both engines before deciding that a brand's website is or is not part of the evidence behind its recommendations. The difference is descriptive because the engines cite different source sets.

The brand-owned rate was 26.0826% for brands ranked #1, 28.3579% at #2, 27.8651% at #3, 27.0618% at #4 to #5, and 27.6991% at #6 or lower. The full range was 2.2753 percentage points.

An owned citation is not reserved for the winner. Who owns the source and where the brand ranked answer different questions, so a citation audit should record both instead of treating a link to the brand's own site as proof of a top placement.

Among the displayed industry groups, Professional Services had 10,885 brand-owned links among 27,411 eligible links, or 39.7103%. Education was at 32.3953% and Artificial Intelligence at 30.3785%. Consumer Goods was at 16.9674% and Consumer Electronics at 10.5798%. Every displayed group had more than 10,000 eligible links.

The 3.75-to-1 range between Professional Services and Consumer Electronics changes the baseline a team should expect. Use the category baseline before labeling a brand's owned-citation rate high or low. These groups cover the prompts that had a main industry assigned in this observed cut.

Among brands with at least 500 eligible links, LinkedIn had the largest owned-link count at 2,341 of 9,834, or 23.8052%. Soar followed with 1,316 of 2,726, or 48.2759%, and Bright Data had 1,188 of 3,356, or 35.3993%. The table is ordered by owned-link count, not rate.

Volume and share identify different work. A high owned-link count shows where a brand's site already supplies substantial evidence. A high share shows dependence on that site. Review both before choosing whether to improve owned pages or broaden independent coverage.

Deeper pages received 195,080 of the 288,629 brand-owned citation links, or 67.5885%. Homepages received 93,549, or 32.4115%. The deeper-page group contained 103,721 distinct pages, compared with 27,805 homepages.

The opportunity with owned content sits at the page level. Keep the homepage accurate, but audit the product, documentation, guide, and other deeper URLs that AI actually cites. We sorted pages by URL depth only; we did not judge what kind of content each page holds.

The current engines produced 1,372,702 brand-to-page citation links in all. We excluded 11,772 without a page URL, 300,413 without a direct ranked brand mention in the same answer, and four without a known primary website domain. That left 1,060,513 links as the base for every rate here.

Exact-domain matching produced a 25.9856% owned rate. Adding documentation and other subdomains on the same registered domain produced the 27.2160% headline. Matching any known domain for the brand produced 28.0042%. The narrow spread shows that the conclusion does not depend on one ownership rule. This method tests who owns the source, not whether the page supports every claim or caused the recommendation.

The observed cut shows that brand-owned pages supply more than one in four citations attached to ranked recommendations, while third-party pages supply nearly three in four. The engine, industry, brand, and page-depth cuts show why one overall score is not enough.

Track three things for every important recommendation: the brand's rank, who owns the cited domain, and the exact cited page. Review ChatGPT Search and Google AI Mode separately. Keep product and documentation pages current, then audit the independent pages that supply the rest of the evidence. Recheck the same prompts after a meaningful page change before crediting movement to that change.