30.4% of ecommerce brands' top cited domains are their own
17.7% the SaaS figure, lowest of the three verticals
15 brands analyzed per vertical

What the study measured

Aleyda Solis took 15 leading brands in each of three verticals, SaaS, ecommerce and finance, and pulled the top 10 cited source domains for each brand across Google AI Mode, Gemini and ChatGPT using Semrush Enterprise data. Every cited domain was then classified by ownership type, which turns a list of URLs into a picture of where an AI answer sources its opinion of a brand.

The result held in every vertical: the brand's own website is a minority of the top cited source mix. SaaS split 82.3% external to 17.7% owned, finance 79.4% to 20.6%, and ecommerce 69.6% to 30.4%.

Solis describes the owned site as an "on-page corroboration base" rather than the main event. The store states product, policy and documentation facts accurately and accessibly; social, community, publisher, review, marketplace and competitor sources then validate, demonstrate, compare or challenge those facts. Both layers are doing work, and only one of them is on your server.

Share of top-10 cited domains the brand owns
Ecommerce
Finance
SaaS

The gap between verticals

Ecommerce sits roughly thirteen points above SaaS on owned share. The study reports the difference without attributing a cause, so anything about why is inference rather than finding.

What the number does support is a correction to how the headline usually travels. "Your own site barely matters" is not what this says for a store. Nearly a third of the cited mix is the store itself, which is the largest owned share of the three verticals measured, and it is exactly the portion that responds to the on-page work an ecommerce team already knows how to do.

The remaining two thirds are the part that does not respond to it. Those citations sit on domains somebody else runs, and which domains they are shifts by prompt.

What to do about it

Pull the cited domains for your own prompts first

Solis recommends identifying which specific sites influence a target prompt before choosing PR or social targets. A vertical average tells you the shape of the problem, not the addresses. Take the ten or fifteen questions your buyers actually ask, run them, and write down which domains the answers lean on. That list is the brief.

Keep the on-page layer accurate enough to corroborate

The corroboration framing puts a specific job on your product, policy and documentation pages: be the accurate, accessible source the external mentions get checked against. Stale specs, missing shipping and returns detail, and product facts that disagree with your feed undermine the third-party work rather than supporting it.