About
A standard for a new shop window
People used to find brands through search engines. Increasingly, they ask an AI assistant — ChatGPT, Gemini, Claude and Perplexity— “what's the best in this category?” or “where should I buy this?” The answer they get, and whether your name is in it, now shapes real buying decisions.
Yet most brands have no idea how they appear in those answers. The AI Visibility Index exists to measure it — rigorously, transparently, and across providers rather than any single one. We give every brand a single AVI score, show exactly why it is what it is, and lay out how to improve it.
We're deliberately vendor-neutral. A number that's only meaningful inside one AI product isn't a standard. Ours measures across the assistants people actually use, with a published methodology and confidence intervals — because a standard has to be trusted to be worth anything.
We started with UK specialty retail — cigars, spirits and the enthusiast brands where being recommended genuinely moves a sale — and we measure ourselves by the same index we sell. Our own reading is public, on the same terms as everyone else's.
How we measure
An AVI Score has two halves. Outcome is what the assistants actually say: we put the questions real buyers ask to each engine, many times over, and count how often a brand is named, recommended and cited — against every rival in its category, so the result is a share of a defined field rather than a number in a vacuum. Readiness is whether a site is built to be found: crawlable by AI agents, readable without JavaScript, described in structured data, corroborated as one unambiguous entity across the web.
The rules that keep the number honest are stricter than they might first appear. A simulated reading is never published as a measurement — a seeded score about a real business is a fabrication, and nine of them were quietly deleted from our own leaderboard the day we understood that. A brand that has not been measured is listed without a rank, because not being measured must never beat being measured honestly. A site that refuses our crawler is shown as refusing, not silently skipped. And every reading names the engines it actually ran on and the version of the question set it sat, so two scores are only ever compared when they are genuinely comparable.
The index its founder can lose on
The first brand this instrument was ever pointed at is our own: City Cigars, the founder's online cigar shop, serves as the index's permanent test subject — measured first, before anyone else's money is spent. In our first full category report it finished last of the five brands measured, and we published that, unadjusted, in our own report. That is not modesty; it is the entire trust model. An index you can believe about your competitors has to be one its own founder can visibly lose on.
Our first category report, on UK cigar retail, judged 1,479 AI answers against a 39-merchant field whose boundary is a published criterion — so when someone asks “share of what?”, the answer is a rule anyone can check, not a list we happened to assemble.
Who's behind it
The AI Visibility Index is built by Eli, Founder — an operator in UK specialty retail who kept asking how often AI assistants actually recommend a brand, found no rigorous answer, and set out to measure it. The index is independent and vendor-neutral: we sell measurement and the work of improving it, not placement, and no brand can pay to change a reading.