Does AI cite brand homepages?

A brand-owned citation link joins a ranked brand recommendation to a cited page on that brand's primary website. Of 289,494 such links, 195,720 pointed to a URL below the homepage. The deeper-page rate was 67.6076%. Homepages received 93,774 links, or 32.3924%.

This result narrows the common assumption that a brand-level domain citation is enough to understand owned visibility. Ahrefs found that article formats dominate Google citation patterns while core-page use varies by search surface. This study measures a different layer: the exact owned page attached to a ranked brand recommendation across ChatGPT Search and Google AI Mode.

Google AI Mode pointed below the homepage for 129,523 of 140,650 brand-owned links, or 92.0889%. ChatGPT Search did so for 66,197 of 148,844 links, or 44.4741%.

One aggregate rate hides two different source patterns. A page-level citation audit should separate the engines before deciding which owned URLs need work. The difference is descriptive because each engine cited a different mix of brands and pages.

Brands ranked #1 received 63,952 deeper-page links among 84,754 brand-owned links, or 75.4560%. The rate was 70.1089% for ranks #2 to #3 and 58.7189% for ranks #4 or lower.

A homepage link is not a proxy for recommendation strength. Rank and page depth describe different parts of the answer, and this cut does not show that a deeper page caused the higher placement.

URLs with two path segments supplied 100,177 of 289,494 brand-owned citation links, or 34.6042%. Homepages supplied 93,774 links, or 32.3924%. One-segment URLs supplied 45,436, and URLs with three or more segments supplied 50,107.

Path depth gives teams a repeatable way to separate homepages from deeper pages. It does not identify the page's subject, quality, or template, so the exact URL still needs review.

The URL-path rules identified 62,743 blog or editorial links and 23,866 product or solutions links. Documentation or support paths supplied 3,448 links. Another 102,765 deeper links did not match those named path groups, and homepages supplied 93,774.

Editorial and product pages both matter, but a path label is only a repeatable rule applied to the URL. It does not prove that every URL contains the content its path suggests. Teams should inspect the leading pages before assigning content work.

Consumer Electronics had 984 deeper-page links among 1,134 brand-owned links, or 86.7725%. The rate was 70.0637% in Information Technology, 69.0807% in Software, 65.2967% in Financial Services, 59.1632% in Sales and Marketing, and 48.2205% in Gaming.

A brand should be compared with its market before its page mix is labeled unusual. These categories describe the prompts present in this observed cut, not every buyer question in each industry.

Among the displayed brands, Stripe Billing received 320 deeper-page links among 321 brand-owned links, or 99.6885%. Jira Software received 412 of 415, or 99.2771%. At the other end, Linear received 33 of 348, or 9.4828%, and Peec AI received 35 of 292, or 11.9863%.

The leaderboard identifies brands whose owned evidence is concentrated on deeper pages or homepages. It does not show whether the cited page is accurate, whether the engine followed the link, or whether the page caused a recommendation.

Atlassian's Jira page received 217 links across 51 prompts. Stripe's Billing page received 161 across 45 prompts. A New Relic observability-tools article received 117 across 23 prompts, and an Overstory vegetation-management article received 113 across 23 prompts.

Repeated citations reveal individual pages worth auditing, but the distribution remains broad. The ten most-linked deeper pages supplied 3,284 of 289,494 brand-owned links, or 1.1344%. Marketers should start with recurring pages and then retain a long-tail review.

We began with 1,390,786 links connecting brands and cited pages in answers from the two measured engines. We excluded 11,839 without a usable page URL, 314,601 without a direct ranked brand recommendation in the same answer, four without a known primary website, and 774,848 links to third-party sites. That left 289,494 links, counted once per answer, brand, and page.

A homepage had no path after the website hostname. Every other URL was a deeper page. Repeated references to the same page for the same brand in one answer counted once. Matching the exact primary hostname produced a 66.6157% deeper-page rate; matching the primary registered website, including subdomains, produced the 67.6076% headline. The conclusion does not depend on the broader rule.

Deeper pages received 195,720 of 289,494 brand-owned citation links, or 67.6076%. Google AI Mode's deeper-page rate was 92.0889%, while ChatGPT Search's was 44.4741%.

Record the recommendation rank, cited URL, page owner, and engine for every priority prompt. Keep the homepage accurate, then audit deeper product, editorial, documentation, and uncategorized pages. Compare each brand with its industry baseline. Recheck the same prompts after a material page change before attributing recommendation movement to that change.