Visibility score

What it measures#

The visibility score answers the most basic question in AI search: when buyers ask the questions you care about, does your brand come up at all?

It is a coverage metric. It counts prompts, not mentions, so a prompt where ChatGPT names you five times counts the same as one where Perplexity names you once at the bottom of a list. What matters is whether the prompt is “won” (you appear) or “missing” (you don’t).

How it is calculated#

• Prompts where your brand was mentioned counts distinct prompts with at least one brand mention of your brand, in any response, from any engine, inside the date range and filters.
• Tracked prompts is every prompt that matches the current filters (country, tag, engine).
• Your brand is matched on the brand name and domain you set when you added the domain, including normalized variants.
• The result is rounded to one decimal place.

Worked example#

You track 40 prompts across ChatGPT, Perplexity and Gemini for the last 30 days. Your brand is named in at least one answer for 14 of them.

14 ÷ 40 × 100 = 35.0%

The other 26 prompts appear as not-ranked prompts: the list to work on first.

The change indicator#

Next to the score the dashboard shows the change against the previous period of the same length. For “last 30 days” that is the 30 days before it. A move from 35.0% to 42.5% displays as +7.5 pts: percentage points, not percent.

Visibility vs. neighbouring metrics#

Read them together. High visibility with low share of voice means you show up everywhere but competitors are named more often. Low visibility with a good average rank means that where you do appear you lead, but you are missing from most prompts: a coverage problem, not a reputation problem.

What moves it#

• The prompt set. Adding 20 prompts where you are not yet mentioned lowers the score immediately, even though nothing about your real visibility changed. Compare periods with the same prompt set, or filter by tag.
• Engines and countries in the filter. Engines differ a lot. Filtering to one engine often explains a swing that the blended number hides.
• Pages that answer the prompt directly. Engines that search the web cite pages that answer the question in plain language. See prompt coverage for which fan-out questions none of your pages answer.
• Being on the pages engines already cite. If a review site or listicle is cited for a prompt and you are not on it, getting listed there often moves visibility faster than publishing your own page. Start from Sources.

Related#