Source analysis is the evaluation of which websites AI systems draw on as sources for the questions in a prompt set: your own and, above all, those of others. It answers the question: on which third-party sites do I need to be present because they are cited often? It is thus the link between measurement and off-page measures.
How does source analysis work?
From every answer to every prompt, the cited URLs are recorded and grouped by domain and page type: listicles and comparison articles, trade media, forums and communities, review platforms, Wikipedia, vendor sites, YouTube. The frequency per domain and per prompt category yields a picture of where the evidence for a topic area comes from. In addition, it is checked whether your own brand appears on these sources, and how it is described there.
The result differs strongly by question type: trade articles and Wikipedia dominate for research questions, listicles for comparison questions, review platforms and forums for trust questions, and the vendor site itself for brand-specific questions.
Why does source analysis matter for AI visibility?
Because for the decisive questions (recommendation, comparison, trust) AI systems predominantly cite third-party sites. A brand that does not appear on the twenty most-cited sources in its field can optimise its own website as much as it likes and remains invisible in these answers. Source analysis makes visible where the work actually lies and prevents effort from flowing into your own site where presence on third-party sites is needed.
What does this mean for your website?
Carry out the source analysis together with the baseline measurement and repeat it regularly; the cited sources change. Derive from it a target list for Digital PR, listicle placements, review profiles and community presence. And check on every relevant third-party site whether your brand is described correctly and consistently there.