Research into how AI systems cite sources for SaaS-related buyer questions is beginning to show a consistent pattern: source selection is concentrated, contextual, and dependent on the platform and prompt. Analyses associated with Kevin Indig and separate B2B SaaS research both suggest that the domains and content formats surfaced in AI answers can vary substantially by buyer stage. That matters for software companies treating AI visibility as a new distribution channel, but it also calls for care when interpreting individual datasets.
The narrow study description circulating around US ChatGPT citations in December 2025, four SaaS buyer-journey stages, and unique cited domains appears directionally consistent with this research theme. However, the supplied evidence does not tie that full methodology cleanly to one published dataset. The most accurate conclusion is that credible studies exist, while their scope, metrics, and engine coverage should not be merged into a single set of findings.
What the citation research indicates
Kevin Indig's analysis of how AI picks its sources examines patterns in AI citation behavior. The research, as summarized in the supplied material, points to a concentration of citations among a relatively small group of domains and to variation across AI engines and content formats. Search Engine Land's coverage of Indig's dataset similarly reports that roughly 30 domains account for a large share of citations, while the balance between first-party and third-party sources changes by platform and prompt type.
