External Sites Hold 70 to 82 Percent of AI Search Citations

Independent SEO practitioner Aleyda Solis has published a cross-vertical synthesis showing that a brand’s own website supplies only 17.7 to 30.4 percent of the domains large language models cite when answering commercial queries. The remaining 69.6 to 82.3 percent, depending on the vertical, comes from social platforms, competitors, marketplaces, news outlets and review sites the brand does not control. Solis published the analysis Sunday on her own site. It synthesizes a SaaS citation study she released in July, an ecommerce study from May, and new finance-vertical data, paired with a fresh practitioner poll.

The split by vertical shows the same pattern with different proportions. SaaS brands’ own pages account for 17.7 percent of top cited sources against 82.3 percent external, of which 45.7 percent is social and community platforms. Ecommerce runs 30.4 percent owned against 69.6 percent external, with 36.6 percent of that external share sitting on competitor and marketplace domains. Finance sits at 20.6 percent owned against 79.4 percent external. News and review coverage accounts for 19.4 percent of that finance total.

Solis built the dataset from her own analysis of Semrush Enterprise data, taking 15 leading brands in each vertical, pulling the ten domains most often cited for each, and doing it separately across Google AI Mode, Gemini and ChatGPT. That makes this a single-sourced study, worth naming plainly. Solis is an independent practitioner, not a vendor selling an AI-visibility product. That is a different reliability profile than the vendor-run studies this brief has tracked in recent weeks, and readers should weigh it accordingly. The findings have not been independently replicated.

One additional detail complicates a simple owned-versus-external read. Across the three verticals, brand-owned domains claimed roughly one in ten of the 150 tracked citation positions, yet those few placements carried a disproportionate share of the overall citation weight, 20.6 percent in finance alone. A brand’s own page, in other words, is scarce in the source mix but punches above its weight when an AI system actually selects it. That argues for treating owned-page accuracy as a precision problem, not a coverage problem: the goal is not more pages, it is pages an AI system can verify without hesitation.

The clearest actionable finding is platform divergence, and it should not be flattened into one generic AI-search playbook. Google AI Mode, Google’s conversational search experience, is consistently social-led: community and creator platforms supply 74.6 percent of SaaS citations there, 41.8 percent in ecommerce and 47.6 percent in finance. ChatGPT leans toward written evaluative content instead. News and review sources climb to 34.2 percent of ChatGPT’s finance citations, a share AI Mode does not approach in that vertical. Gemini is mixed and vertical-dependent. Competitor domains account for 44.7 percent of its ecommerce citations, closer to ChatGPT’s pattern than to AI Mode’s.

Solis’s recommendation is to sequence the two efforts rather than choose between them: “build the on-page evidence base, then use source analysis to prioritize the relevant third-party environments.” For SaaS teams that means budgeting toward video, Reddit and YouTube presence. For ecommerce it means auditing marketplace, retailer and competitor listings. For finance it means securing accurate placement on comparison and review sites.

One supporting data point should be read with caution. Solis ran a LinkedIn poll among her own followers that had drawn 223 votes and was still open when she captured the results, an informal subsample of her professional network rather than a scientific survey. Fifty-three percent of respondents said they invest in both digital PR and social media for AI search visibility, 21 percent said digital PR only, 8 percent said social only, and 16 percent said neither.

The practical implication follows directly from the citation math. If 70 to 82 percent of the citations shaping AI visibility sit on pages a brand does not own, on-page optimization is necessary and no longer sufficient by itself. Digital PR and community seeding move from discretionary add-ons to a budgeted acquisition channel. Sizing that channel means matching the vertical’s citation mix and the platform your buyers actually use, not applying one playbook everywhere. Pull your own citation panel by platform before deciding whether the next budget line goes to Reddit and YouTube or to comparison-site placements.

Aleyda Solis published this analysis, including the SaaS, ecommerce and finance citation data and LinkedIn poll, on her own site, aleydasolis.com, on August 2, 2026.