What Is AI Visibility Readiness (AVR)? Definition and Formula

In July 2026 we ran a controlled measurement of citability.dev against two AI engines, and one result was worth the whole exercise: when we asked "what is AI Visibility Readiness (AVR)," neither Claude nor OpenAI cited us. We coined the term. We publish the open-source framework for it. And on the day we measured, the engines that shape buyer perception could not point a single link back to the domain that owns the concept. That is the gap AVR is designed to make visible, and this post exists partly to close it for our own site.

AI Visibility Readiness is a single score for whether AI answer engines can find, trust, and cite your brand. It sits one level up from AI citability, which is one of its three axes. Where citability asks the narrow question "does the engine link to you," AVR asks the fuller question a buyer's journey actually runs through: are you mentioned at all, are you recommended when someone asks who to hire, and are you cited with a link they can click.

The Formula#

AVR = mean(visibility%, recommendability%, citability%).

Three rates, averaged. Each is a countable proportion over a fixed query set, which is the entire point: the number is reproducible by anyone who runs the same queries, not a proprietary index you have to trust.

• Visibility is the mention rate. Of all answers, how often does the brand name appear anywhere in the text.
• Recommendability is the recommend rate on buyer-intent queries. When the engine is asked "which tool should I use" or "who can I hire for this," how often does it name the brand.
• Citability is the cited-with-link rate. How often a working link to the brand's own domain appears in the sources.

A brand mentioned in half of all answers, recommended in none of the buyer queries, and cited a quarter of the time scores mean(50, 0, 25) = 25. The three numbers are kept apart on purpose. They fail independently. The most common failure we see is a brand that is mentioned often and cited almost never, a pattern we call a dark mention: the model knows you exist but sends the click somewhere else.

Why Three Axes and Not One#

Most tools in this category report a single "AI visibility score." In practice that number is usually a mention rate wearing a bigger name. The problem is that mentions do not pay. A buyer reading an AI answer does not convert on your brand being named in a sentence; they convert on the link they can click and on the engine recommending you for the job they are trying to hire for.

Collapsing everything into one score hides exactly the axis that is broken. When we split citability.dev's own July 2026 run apart, the composite AVR was 32.6 percent, but that average concealed a real split: on the neutral measurement, the recommendability axis was where the domain was thinnest, not the mention axis. You cannot fix what the average hides. AVR keeps the three axes visible so the weak one is diagnosable.

There is a second reason to insist on the recommendability axis: query intent. The Conductor 2026 benchmark found that informational queries are a shrinking share of AI answer surfaces while commercial and navigational intents carry real weight. If your measurement only asks informational questions ("what is X"), recommendability is structurally unmeasurable and your AVR is quietly capped. A buyer-intent query set is not optional; it is the only way the recommendability number means anything.

A Measured Example, With Its Limits Stated#

Numbers without a confidence label are marketing, so here is the honest version. In a July 2026 competitor run over an identical buyer-intent query set, an established AI-visibility tool scored AVR 72, while two newer entrants scored 35 and 20. That is a real, directional spread on the same queries and the same engines. It is also a low-confidence result: a single round, a small query set, no confidence interval, and the subscribed chat engines showed self-subscription bias that a neutral engine (Perplexity) did not share. On one buyer query the subscribed engines favored one tool while Perplexity cited a different one entirely.

The lesson is baked into the method. Report the query count. Report the confidence. Treat a neutral retrieval engine as the signal and your own logged-in accounts as suspect. An AVR score quoted without those guardrails is not measurement, it is a flattering screenshot.

What AVR Tells You to Do Next#

The value of splitting the score is that each axis points at a different fix.

• Low visibility (rarely mentioned) is an authority and coverage problem. The engines do not associate your brand with the topic yet. This is where entity structure, original data, and third-party coverage move the number.
• Low recommendability (mentioned but never recommended) is an off-site trust problem. The engine knows you but does not vouch for you when a buyer asks who to hire. This is won on the sources the engine already trusts, not on your own pages.
• Low citability (mentioned but not linked) is a structural and formatting problem. Your content is not the cleanest extractable source for the claim, so the engine names you and links someone else.

Our own site is the worked example. citability.dev has partial visibility, a genuinely weak recommendability axis, and, on the day we measured, zero citation for its own coined term. The framework refuses to flatter its owner, which is the only reason its numbers are worth anything. For the traditional-search half of this same story, and why ranking on Google no longer guarantees an AI citation, see the companion write-up on answer engine optimization.

AVR is not a grade you pass once. It is a repeatable instrument that tells you, per engine and per intent, exactly where you are invisible and which of three levers to pull. The score is only as honest as the query set and the confidence label attached to it, so we publish both.