The five numbers behind your AI Shortlist Score

One score, five ways to read it

A single number out of 100 is easy to remember and easy to compare against last month. On its own, it's almost useless for deciding what to do next. VISIBLE's AI Shortlist Score is built by classifying every answer an assistant gives — across a fixed, frozen set of Money Prompts — on a five-step ladder, and then reading that same underlying classification back through five different buyer-facing questions. A score that dropped from 46 to 31 tells you something changed; it doesn't tell you what, until you look at the five numbers underneath it.

A mention is not a recommendation.

VISIBLE

That line is the whole design principle behind the score. An assistant naming your brand in passing and an assistant actively putting your brand forward are two different outcomes, and a measurement that can't tell them apart isn't measuring anything a buyer would recognise as visibility. So every answer gets sorted onto one rung of the same ladder before anything is counted.

The five metrics

That classification is what makes five separate buyer-facing numbers possible, each one reading the same underlying answers for a different purpose:

• AI Shortlist Score — Your overall commercial recommendation health across the buyer questions measured.
• Share of Shortlist — Your share of considered and recommended positions versus tracked competitors.
• Recommendation Rate — How often AI actively proposes your brand rather than merely naming it.
• Discovery Win Rate — How often you enter the shortlist when the buyer does not already know your name.
• Brand Truth Score — How accurately and consistently AI understands the material facts about your brand.

Read plainly, those five split into three kinds of question. The Score is a single-glance summary. Share of Shortlist and Recommendation Rate both describe competitiveness — one relative to rivals, one about how confidently you're put forward once you're in the conversation at all. Discovery Win Rate is the harder, more valuable one: it isolates what happens when a buyer doesn't already know your name, which is the moment most AI visibility tools quietly skip because it's the one their own instrument can't easily fake. Brand Truth Score sits apart from all of them — you can be recommended often and still be described inaccurately, and neither of the other four numbers would catch that on its own.

Which one to chase first

The five numbers only earn their keep once they're read against each other. A brand with a reasonable Recommendation Rate but a weak Discovery Win Rate is winning the questions where a buyer already half-knows them and losing the ones where they don't — a discovery problem, not a trust problem, and it usually means the brand's evidence simply isn't reaching the sources an assistant checks on a cold, category-first question. The reverse pattern — decent Discovery Win Rate, weak Recommendation Rate — means the brand keeps getting found and then talked past: named, considered, and set aside for someone the assistant trusts more. Those two brands can land on the same overall AI Shortlist Score and need entirely different fixes.

A sample result

None of the figures below belong to a real brand. They're the sample result VISIBLE shows to illustrate what a first measurement typically looks like — the shape of the number, not a claim about any actual company.

• AI Shortlist Score: 31 / 100
• Money Prompts won: 3 / 12
• First-choice recommendations: 1 / 12
• Questions competitors own: 7 / 12
• Strongest competitor: 67 / 100

That gap between 31 and 67 is doing less work than it looks like on its own. The more useful detail sits inside the 12 Money Prompts behind it: winning 3 of them outright but landing as the buyer's actual first choice on only 1 is a Recommendation Rate story more than a Discovery Win Rate one — this brand is getting into the conversation and then losing it at the final step, more often than it's failing to show up at all. Competitors owning 7 of the 12 outright, meanwhile, is the clearest single reason the overall Score sits closer to 31 than to the middle of the range — more than half the category's Money Prompts are going somewhere else entirely.

How the score combines

It's worth being precise about how this fits together, because the five metrics above and VISIBLE's audit engine are easy to conflate and actually describe two different things. The engine scores every measurement on a fixed, published composite — VISIBLE = 100 × (0.35 × Presence + 0.30 × Recommendation + 0.20 × Authority + 0.15 × Accuracy) — with each of those four dimensions averaged equally across every engine that measured it. That formula is the audit service's own; it isn't something invented for this article, and it's the one actually verified in the codebase (`aev/config.py`, `VISIBLE_FORMULA`). The five buyer-facing metrics above aren't a second, competing formula sitting next to it: they're how that same underlying set of measurements reads once it's turned toward a buying decision rather than an internal score — one overall-health read, one competitive read, one confidence read, one discoverability read and one factual read, all drawn from the same evidence. VISIBLE doesn't publish a weighting across those five, because there isn't one to publish; they answer five different questions, not five shares of one number.

What's kept separate

None of the five numbers include how crawlable your site is, whether it has the right structured data, or whether its pages render without depending on client-side scripts. Technical readiness is real, and it's one of the reasons a brand with strong facts and a decent reputation can still score poorly — if the pages carrying that information are hard for an assistant to read in the first place — but it's reported on its own, with its own checks, rather than blended into a number that's supposed to describe recommendation. A technically pristine site with nothing distinctive to say would otherwise drag the composite up for the wrong reason, and a messy but well-regarded brand would get penalised for one.

What the score doesn't promise

None of the five numbers guarantee a ranking, and none of them are a prediction. AI answers are dynamic — the same question can return a different shortlist a month later — so every one of these figures is a measurement of a fixed set of Money Prompts at a point in time, repeatable on a schedule so movement becomes visible rather than anecdotal.

The score also doesn't publish a guess at total AI-conversation volume for your category — that figure can't be verified and would only manufacture a false sense of scale.

Free Snapshot vs the $99 Intelligence Audit

A free Snapshot shows an AI Shortlist Score preview and a Share of Shortlist preview, alongside three questions you're losing and one visible Brand Truth issue — enough to see the shape of the problem across 8 Money Prompts and up to 3 competitors. The full Intelligence Audit repeats all five metrics against 30 Money Prompts and five tracked competitors, with the evidence behind each number inspectable rather than summarised, plus a Brand Truth Gap register and a 90-day action plan. The five metrics themselves don't change between the two; what changes is how many Money Prompts and competitors they're measured against, and how much of the underlying evidence you can actually see.

Once you know which of the five numbers is weakest, how competitors win the recommendation or where the facts went wrong usually explains why, and Share of Shortlist goes deeper on reading competitiveness specifically. Seeing your own five numbers starts with the free Snapshot; the full breakdown, competitor by competitor, is in the Intelligence Audit.