How to read your first LLM Visibility Report
The report is built so the first screen answers the question and the next fourteen show the evidence. Here is how to read it in the order it was written — and what each number can and cannot tell you.
Start with one number, then distrust it a little
The first thing on the first screen is AI Visibility: a combined measure of how often and how prominently the brand appears in relevant AI answers, across all three engines. In the sample report for Vega Skates — a fictional brand we use in every example — it is 42%.
Treat that number the way you treat a temperature. It tells you where the brand is today, and it only becomes a story when you have a second reading. The four tiles next to it are the parts it is made of: mention rate, recommendation rate, average position and citation rate. When the headline number moves, one of those four moved it, and the tiles tell you which.
Mentioned is not recommended
A brand can be named in most answers and recommended in few of them. An assistant that lists five brands and then says which one it would pick has mentioned all five and recommended one.
The report keeps the two rates apart on purpose. A large gap between them is a positioning problem, not a visibility problem: the engines know you exist and do not have a reason to pick you. That usually points at product pages that describe the product without saying who it is for.
Three engines, three opinions
Sections four to six give each engine its own screen because they rarely agree. ChatGPT, Gemini and Perplexity retrieve from different sources, weight them differently and phrase their answers differently. A brand that Perplexity cites in every answer can be absent from Gemini entirely.
Read each screen for the same three things: the prompts where the engine names you, the prompts where it does not, and the example answer quoted word for word. The quote matters more than it looks — it is the sentence a buyer actually sees, and it is the fastest way to explain the whole report to someone who has not read it.
The heatmap is your to-do list
Section seven puts every tracked prompt against every engine, one tile each: dark when the brand was mentioned in the majority of runs, light when it was not. It is the most-forwarded screen in the report, and it is also the most practical one.
- A row of dark tiles is a prompt you own. Leave it alone.
- A row of light tiles is a question the engines cannot answer with your brand — usually because no page of yours answers it either.
- A mixed row is a prompt where one engine found a source about you and the others did not. Look at which sources that engine cited.
Competitors: the gap to the leader is the story
Section eight shows the brands the engines named in the same answers, with each brand's share of answers. Your brand is highlighted. The number to take away is not your share; it is the distance between you and the brand at the top.
In the Vega Skates sample the leader is named in about half of all answers and Vega in a little over a third. That gap is made of specific prompts — the ones in the heatmap where the leader has a dark tile and Vega has a light one — and the opportunities section lists them.
What to do in the first month
The last sections turn the numbers into work. The order they suggest is deliberate.
- Write or rewrite the pages behind the empty heatmap rows. One prompt, one page that answers it in plain language.
- Fix the product pages: who the product is for, what it costs, what makes it different. That is what turns a mention into a recommendation.
- Get named on the sources the engines already cite for your category. The sources section tells you which ones.
- Order the second report. Nothing in the first one is a trend until you have it.