Metrics

Visibility

How Ranqo calculates Visibility, the share of AI answers that name your brand, and why it is weighted by engine.

Visibility answers the first question a brand asks about AI search: when buyers ask about your category, does AI name you?

Visibility: The share of your tracked prompts where AI mentions your brand.

Definition, in the same words as the dashboard's hover text.

Visibility is the share of AI answers to your tracked prompts that name your brand, with each engine weighted as described below. One answer is one prompt asked on one engine in one run. The hover text on the Visibility card calls it the share of your tracked prompts, but the count is of answers.

One answer, one verdict

Each run asks every tracked prompt on each engine it is set to. Each of those answers gets one verdict: it names your brand, or it does not. An answer names you when it uses your brand's name or one of your aliases to refer to you. A mention that appears only after Ranqo reformats the answer, and not in the engine's original text, is withdrawn. How often you are named inside one answer does not matter; an answer that names you five times counts once. See How Ranqo measures for how an answer is read.

On one engine, visibility is a plain rate:

Calculated on each engine separately

Weighted by engine

Across engines, Ranqo does not pool every answer into one count. It calculates visibility on each engine separately, then combines them with fixed engine weights that Ranqo sets, broadly by each engine's reach:

ChatGPT 25%, Google AI Overviews 20%, Claude 15%, Perplexity 15%, Gemini 15%, and Grok 10%

A brand only ever uses the engines it tracks, so the weights are rescaled over those engines to add up to 100%. An engine with no answers in the range you are looking at drops out of the calculation rather than counting as zero.

Weighting keeps the number meaningful when engines disagree. If you are named often on a small engine and rarely on the largest one, a pooled count would overstate your reach.

A worked example

Take a brand tracking two engines, with illustrative weights where engine A counts twice as much as engine B. Each engine answers 20 prompts.

Engine A names you in 12 of 20 answers: 60%.
Engine B names you in 2 of 20 answers: 10%.
Visibility is (60 × 2 + 10 × 1) / 3 = 43.3%.

Pooling the answers would give 14 of 40, which is 35%. That is why the headline number can differ from a raw count of mentions shown elsewhere on the page.

Per engine and per category

Per engine. The Platform Breakdown card shows each engine's plain rate. Hover a visibility figure to see the count of answers behind it.
Per category. The Category Visibility card calculates each prompt category the same way as the headline: per engine first, then weighted. A category holds fewer prompts than your whole set, so its number moves more from run to run.

Each breakdown card marks your overall visibility, the same figure as the Visibility card, with a thin tick on every bar and an Avg caption above the bars, so you can see at a glance which rows sit above or below it.

What the Visibility card covers

The Visibility card on the Visibility page counts the answers from the runs that completed in the date range you pick, and only answers that came back in their own run: a failed query and an answer reused in its place are both left out. See When a query fails.

The change beside it compares the same calculation over the previous range of the same length, and it is a relative change: going from 20% to 25% shows as 25%, not 5 points.
In your first range, when the range before it holds no run, the card reads Stable. There was nothing to compare with.
When the range holds no run, the card shows the last completed run's value with no change. The card does not show that run's date; the trend chart below it does, with an As of note. See Carried values.

Each run also has its own score, calculated over that one run, which you see in Snapshots. A run's own score counts reused answers, so when a run had failed queries it can differ from the Visibility card, even for a range that holds only that run.

How it is displayed

Visibility is shown to one decimal place, and a whole number drops the decimal: 43.3% and 25%. The value is rounded before it is formatted, so a weighted 99.98 shows as 100.

Things that affect the number

Branded prompts. A prompt that names your brand is answered with your brand almost every time. The Unbranded option in the prompt filter on the Visibility page leaves out prompts flagged as branded. Generated prompts are flagged when they name you; a prompt you add yourself is not flagged, even when it names you.
Google AI Overviews. When Google shows no AI Overview for a prompt, that answer counts as one that does not name you.
Failed queries. A run's own score reuses a recent answer in place of a failed query, or leaves the query out. The cards and breakdowns on the Visibility page leave out both. The Tracking tab of the Prompts page and a prompt's insights page count a failed query as an answer that does not name you. See When a query fails.
Long answers. The reformatting reads only the first 8,000 characters of an answer, and its output is what gets scored, so a brand named only after that point counts as not named unless the reformat was discarded. See How an answer is read.
Sampling. AI answers change from run to run even when nothing else has. Small moves are usually noise. See Noise bands.

Visibility asks whether AI names you at all. Share of Voice asks how much of the conversation is yours: your mentions against every qualified brand mention in the same answers. A brand can be named in most answers and still hold a small share when competitors are named more often. Read the two together.