AiRanks asks AI models thousands of real buying questions, such as which CRM is best for a small business, and stores each answer as an immutable observation. From that corpus it derives three things buyers care about: which brands get named, which domains get cited as sources, and which specific pages are linked. Each domain then receives an AIR score from 1 to 10, a visibility decile within the active dataset, with zero meaning not enough data. The published dataset referenced on the site, code-named COBALT, covers roughly 1,446 ranked questions and hundreds of thousands of sampled responses.
What distinguishes the product from typical rank trackers is its statistical posture. Rankings come with bootstrap confidence intervals, so you can tell whether a position change is real or sampling noise, and the site openly documents how observations become scores. It also analyses tracked links inside AI answers, separating ordinary citations from links carrying ad or click-tracking identifiers, which gives marketers a rare view of where paid placements actually surface in AI responses.
Alongside ranking data there is an audit side: a free domain report checks the files AI crawlers look for, including llms.txt, llms-full.txt, ai.txt, ai-plugin.json, sitemap.xml, robots.txt and JSON-LD structured data. For teams trying to be discovered and quoted by AI assistants, this combines competitive measurement with practical on-site fixes, and there is a public API for pulling domains, brands and phrase data programmatically.
