How we measure

The free audit crawls your site once — homepage, up to ~8 pages, robots.txt, sitemap — exactly the way an AI crawler does: no JavaScript execution, honest user-agent (SeeGeoAudit/1.0, documented in our crawler library). Checks are deterministic: the same site produces the same grade. Scores weight crawler access, technical foundation, structured data, content extractability, off-site presence, and entity clarity. When we can't reach or evaluate something, the report says “couldn't verify” — an unreachable domain gets no grade at all rather than a fake F.

AI answers are non-deterministic: the same question asked twice returns different brand lists. Any tool selling you a fixed “AI rank” is selling noise. SeeGeo runs your prompt panel repeatedly (three runs per prompt per scan) across every engine we measure, and reports visibility — the share of readings where an engine's answer named your brand — as a trend over time. Where an answer happens to contain a ranked list, we show your observed position in that one answer, labeled as exactly that. Citations are measured only when an answer actually carried citations; “no citations to check” is reported, not counted against you. Failed API calls are excluded from every statistic — an error is never scored as invisibility.

Daily plans scan daily, but search grounding refreshes weekly: answer text (which drives the visibility trend) changes day to day, while the pages engines retrieve change on the timescale of the web. This is also why we can include every engine on every plan without per-engine pricing games.

Our own site is graded by the same engine (it gave us a C; we published it), our 107-site research runs on the same checks with the full methodology published and the aggregate data on GitHub, and our own weekly visibility numbers are measured by the same scanner our customers get.