How to measure website visibility in LLM answers
Why search rankings are no longer the whole picture
A classic rank tracker records a URL's position for a keyword. An LLM works differently: it can name several brands, build a recommendation from multiple sources or omit a page that ranks well in Google.
So 'where does my site rank?' becomes two questions: does the brand appear, and what role does it get — recommendation, alternative, source of a fact or passing mention?
Start with questions, not a keyword list
A prompt should sound like a real customer request: 'which services should I choose for…', 'where can I buy…', 'compare…', or 'what is best for a company with…'. These formulations reflect the user's task better than a single keyword.
Split the library into category, product, comparison, problem, region and purchase intent. Tags later show where visibility grows and where the brand disappears.
Which metrics are worth tracking
Start with the share of prompts that mention the brand, relative position in recommendations, Share of Voice and the number of answers containing a link to your domain. Each metric answers a different part of the visibility question.
Do not hide providers inside one anonymous average. ChatGPT, Perplexity, Google AI and other systems use different interfaces and sources, so you need both the overall trend and the provider-level difference.
Why one measurement is not enough
A generative answer is probabilistic and can change with a model update, source mix, region or prompt wording. One screenshot captures an event, but cannot tell you whether it is durable.
Repeat the same prompts and save runs. Comparing one method across dates turns an observation into a signal: the brand is truly losing visibility, or the answer simply took another form.
What to do after measuring
If the brand is absent, audit relevant pages: is the direct answer present, are facts and structure clear, and are supporting sources available? If it appears but trails competitors, compare completeness and evidence.
LLM Control exposes the gap and its trend. For a page-level diagnosis, use AI Readiness to see how the content performs for retrieval, extraction and citation.
Want to check this in your market?
LLM Control regularly collects AI responses, brand positions, competitors and cited sources for your prompt library.
Is your own content set up for stories like these?
Run a free AI Readiness check — retrieval, extractability, schema.org signals, and a prioritized rewrite brief, scored the way an AI assistant actually reads your page.