30-second rundown
Key learnings:
- Outside sources dominate: AI systems usually learn about SaaS brands from third-party websites, not only from the brand’s own site.
- The platforms hunt differently: ChatGPT favors written comparisons, communities, and publications more than Google AI Mode, which leans more toward video and social sources.
- Citations and visits are different: Measure where AI found the evidence separately from where a visitor landed, because those pages are often different.
This week: Choose five real buyer questions. Ask each one in ChatGPT and Google AI Mode, record every cited source, label it as owned, community, video, review, publication, or competitor, and note where any resulting visit would ideally land.
Coming 3 months: Repeat the same questions monthly and give one owner responsibility for the source map. Improve documentation and guides on your site while PR, community, social, and video teams strengthen the outside sources that each platform favors. Track citations, referral landing pages, and conversions separately.
Aleyda Solis went hunting through the citation swamp with 15 major SaaS brands and came back with the carcass of a comforting myth.
Your website is not the center of the AI universe. It is one dusty building on a highway crowded with reviewers, communities, competitors, videos, and machines that do not agree on where to stop.
1. Most of the Evidence Lives Somewhere Else
Across CRM and sales, project and collaboration, and accounting and finance brands, third-party sources supplied roughly 84 to 93 percent of the citation weight in ChatGPT and Google AI Mode. Owned pages still matter because they establish product facts, documentation, and proof. But betting the entire operation on your own blog is like opening a casino with one chair and no door.
2. ChatGPT and Google Hunt Different Animals
ChatGPT leaned more heavily on peer software sites, communities, tech publications, reviews, and statistics. Google AI Mode leaned much harder on video and social sources, while guides carried more weight among the owned pages it cited. A single distribution plan therefore leaves one machine fed and the other snarling in the parking lot.
3. A Citation Is Not the Same as a Visit
The page that teaches an AI system about a brand is often not the page a person visits afterward. In the three SaaS groups, 22.4 to 67.3 percent of measured AI referral traffic landed outside the URLs found in the citation set. Track the source that influenced the answer, the page that received the click, and the action the visitor took as three separate events, or the dashboard becomes a slot machine wearing a lab coat.
The practical verdict is ugly but liberating.
Keep your own facts clean, then build evidence in the outside places each machine already trusts. The brand that learns to work both sides of this divided desert gets remembered while the rest keep polishing an empty homepage.
