What is an AI visibility score?

What does an AI visibility score actually measure?

An AI visibility score measures two distinct outcomes, not one. The first is whether an assistant names your business in its answer text. The second is whether it cites your website as a source. These come apart more often than owners expect. In a June 2026 check of 80 answers, one tool was named in 16 answers while its domain was cited as a source in only 1. A name reflects reputation the engine picked up elsewhere; a citation means the engine read and quoted your own pages.

A trustworthy score is built from answers to real buyer questions, not from brand lookups. Asking an engine "do you know my business?" measures recognition of a name you supplied. Asking "who should I hire for X" measures whether you surface when a stranger has buying intent. Only the second kind of question produces a score that reflects how customers actually find you, so that is what the score is built on.

Because names and citations are tracked separately, a single number can hide two very different situations. A business with a strong reputation might be named often yet cited rarely, which tells you the engines trust your brand but are not reading your pages. A newer business might earn a few citations before any mentions, because a clear answer page got quoted before the brand became known. The score keeps both figures visible so you can see which half is working and which half needs the next round of pages.

How is citation share calculated?

Citation share is the core of the score. It is the number of answers that cite your domain divided by the total number of answers measured, counted per engine and then combined. An answer counts toward your citation share when your website appears in that answer's list of sources — being named in the text is a separate metric and does not count here.

Work a small example. Suppose you measure 10 buyer questions across 4 engines, which produces 40 answers. Your domain is cited in 3 ChatGPT answers, 2 Claude answers, 1 Gemini answer, and 0 Perplexity answers, for 6 citing answers in total. Citation share is 6 ÷ 40 = 0.15, or 15%. Per engine it is 3/10 (30%) on ChatGPT, 2/10 (20%) on Claude, 1/10 (10%) on Gemini, and 0/10 (0%) on Perplexity. The combined 15% and the four per-engine figures both go into the score, because a business can be visible on one engine and absent on three.

The per-engine breakdown is the part owners tend to skip, and it is where the useful detail sits. In the example above your combined citation share is a respectable-looking 15%, but the engine split shows you are invisible on Perplexity and weak on Gemini while doing reasonably well on ChatGPT. Averaging those into one figure would hide the two engines where you have no presence at all. Counting per engine first, then combining, keeps that information in view so you know where the next page should point.

The phrase "ai visibility score" is searched about 110 times a month in the US (DataForSEO, June 2026), so the way it is calculated is a question owners are actively asking.

How is the score weighted?

Tenva's published score combines four components on a 0–100 scale. Each component is a ratio measured from the same set of answers, then weighted and summed. Citation share carries the most weight, and the table below shows the full breakdown.

Citations weigh most because a citation is the strongest signal an engine can give: it means the assistant read your pages and chose to quote them as a source. A mention is weaker evidence, since the engine may be repeating reputation it absorbed from third parties without ever reading your site. Share of voice places your result against the competitors the engines actually name, and cross-engine consistency rewards businesses that hold up across all four engines rather than spiking on one.

What does a worked example look like?

Apply the weighting to a real measurement. The June 2026 baseline below scored zero on every component, which is the typical outcome of a first check. With no citations, no mentions, and no presence on any engine, citation share, mention share, share of voice, and cross-engine consistency all evaluate to 0, so the weighted total is 0 out of 100.

To see how the weighting would behave once the number starts moving, picture the same business after a few months of work. Say it reaches a 10% citation share, a 15% mention share, a 20% competitive share of voice, and appears on 2 of the 4 engines for a 50% consistency figure. The weighted total is (0.40 × 10) + (0.30 × 15) + (0.20 × 20) + (0.10 × 50), which works out to 4 + 4.5 + 4 + 5 = 17.5 out of 100. That number looks small, yet it represents real citations on real buyer questions where there were none before.

A zero looks discouraging, but it is the normal starting point and a clean baseline to measure against. The same run also showed competitors were beatable: a name/citation divergence meant one tool's reputation outran its cited pages, which is the gap a new entrant can close with quotable answer pages. Getting cited in AI answers is the work that moves the citation-share component, and getting ChatGPT to recommend your business covers the same job for the largest single engine.

How is an AI visibility score different from share of voice or a Google ranking?

The three measure different things on different units. An AI visibility score measures whether assistants quote and name you in answers; share of voice measures your portion of attention against competitors; a Google ranking measures where a page sits in classic search results. A business can score well on one and poorly on another, so they are not interchangeable. The matrix below sets them side by side.

The practical consequence is that the three numbers can disagree, and that disagreement is information. A page can rank on page one of Google and still never be quoted by an assistant, because the engine could not find a clean answer to lift from it. A page that ranks nowhere on Google can be cited repeatedly, because it answers one specific question in plain language an assistant can quote. Reading the AI visibility score against your Google rankings shows you where classic SEO has already done the groundwork and where the answer pages still need to be written.

Share of voice is one input to the AI visibility score rather than a substitute for it, and the Google rank lives in a separate system entirely. You can read more on which engines behave how in which AI engines cite small-business websites.

What makes an AI visibility score unreliable?

A score is only as trustworthy as the measurement under it, and four failure modes quietly inflate or distort the number.

1. Single-engine checks. A score drawn from ChatGPT alone says nothing about Claude, Gemini, or Perplexity. The engines use different sources, so a one-engine number misrepresents three quarters of where your customers ask.
2. Counting only name-mentions. Mentions are reputation the engine repeats; citations mean your pages were read. A score that counts mentions and ignores citations rewards the wrong outcome and overstates how visible you are.
3. One-run snapshots. Answers vary between runs of the same question. A score from a single run is only a snapshot, and the figure can swing when you re-measure the same questions a week later.
4. No underlying questions or answers shown. A number with no visible questions, answers, or cited sources cannot be checked. If you cannot inspect the answers behind the score, it is a marketing widget rather than a measurement.

A reliable score asks real buyer questions across all four engines, counts citations and mentions separately, repeats the same questions over time, and shows you every answer. The method behind these requirements is documented in Tenva's published AI visibility baseline, and the same standard underpins the free AI visibility check.