The question is usually stated as “how visible am I in AI search?”, and that phrasing hides the choice. A site can be perfectly readable to a crawler and never mentioned; it can be named in answers while the cited page belongs to somebody else; it can rank first in conventional search and be absent from the assistant panel above it. Each of those is a different measurement, and measuring the wrong one is how a team spends a quarter improving a number that does not describe the problem.
This page puts the four measurements side by side. The two that are most often confused with each other have their own pages, linked in the table and again at the end.
What are the four measurements?
Each row is something a reader might buy, build or run by hand. The column that settles most arguments is the last one.
| Measurement | The question it answers | The data it needs | What it cannot tell you |
|---|---|---|---|
| Readiness audit (this site) | Can a crawler reach this page, parse it, and find what an answer would need? | Your URLs fetched once, plus robots.txt, llms.txt and the sitemap; the page response is scored against published rules. | Whether any engine actually mentions or cites you. It measures readiness, not answers. |
| Brand-mention tracking | Is my brand named in AI answers, how often, and in what context? | A question set and access to model responses, a brand name and its aliases, and a way to count appearances over time. | Whether the pages being cited are yours, or why a mention appeared. A mention is not a citation. |
| Rank tracking | Where do my pages appear in conventional search results for a keyword set? | Search result pages for the keywords, country and language you care about, collected on a schedule. | Whether an assistant quotes or recommends you. A rank position and an answer are different surfaces, and the answer can cite nobody. |
| Crawler and server-log analytics | Which AI crawlers fetch my pages, when, and what did they receive? | Your own server or CDN logs: user agent, path, status code and timestamp. This is the only one of the four measured on infrastructure you control. | Whether a crawl became a citation. Crawlers fetch far more pages than any engine quotes, and a 200 in a log is not a mention. |
Which measurement should I start with?
Start with the one whose failures you can fix yourself. That is usually the readiness audit, because it reads pages you control and every failed check names an edit.
- If crawler access is unknown, audit first. It is free here, it reads one URL, and every rule it applies is published with its points on the methodology page. A blocked crawl makes the other three measurements describe a page no engine can read.
- If the pages are readable and the question is presence, you need brand-mention tracking. It answers a different question from a different data source, and the comparison with DataForSEO’s product sets out where the boundary is.
- If search traffic still pays the bills, rank tracking remains a real measurement of the surface your customers still click on. It simply is not a measurement of assistant answers.
- If you suspect a crawler is turned away at the edge, read the logs. Your CDN is the only place a 403 aimed at a bot is visible, which is why the study uses a browser-shaped second request to tell a policy block from a network block.
Can one measurement stand in for another?
No. The four sit at different points in the same pipeline, and each one is blind exactly where the next one starts. A readiness score cannot see an answer. A mention count cannot see the page that was cited. A rank position cannot see a recommendation. A crawl log cannot see a citation.
That is why the useful comparison content here is about choosing between them rather than about declaring a winner. The scanner runs 40 checks across 12 weighted dimensions, and AI crawler access alone is 16% of the score because a page that cannot be fetched cannot be read. The AI context file, by contrast, carries 5% of the score - a real signal, deliberately small. Those weights describe readiness; they say nothing about whether an engine will pick you.
Where do these pages go next?
- LLMention GEO Scanner vs DataForSEO’s LLM Mentions API — the two products whose names are most often confused, compared on data source, unit of result and blind spots.
- Can AI cite you, or is it already doing so? — the conceptual version: why readiness is upstream of mentions, and how to use both in the right order.
- The methodology — every rule, its weight and its pass condition, from the same catalogue the scanner runs.
- Run the audit — the free measurement this site provides, with the failed checks listed against the evidence each one saw.