How we measure whether AI search is actually working in Vienna
Everyone wants to "show up in ChatGPT" now, and plenty of agencies are happy to charge for it while measuring nothing real. AI search — being cited when someone asks an assistant for a recommendation instead of typing into Google — is genuinely worth pursuing. But it is new enough that the metrics are still full of noise, and knowing what to count is the whole game. Here is how we actually judge it, and the numbers we deliberately ignore.
The measurement that matters is citation presence: when a real Vienna-area buyer asks an assistant a real buying question, does your business come up, and in what light? We build a set of the actual prompts your customers would use — "best [your service] near Vienna VA," "who does [your trade] in Oakton," "is [your business] any good" — and we check, on a schedule, whether the major assistants surface you, ignore you, or cite a competitor. That presence-or-absence, tracked over time and tied to specific prompts, is the honest scoreboard. It is slower and less flashy than a dashboard full of green arrows, but it reflects reality.
We also watch what the assistants say about you, not just whether they mention you. In an affluent, research-heavy market like this one, a customer often asks an assistant to compare two or three Vienna options before they ever pick up the phone. If the AI is repeating stale hours, an old address, or a competitor's talking point, that is a problem worth fixing — and it is invisible unless someone is actually reading the answers.
The third real signal is downstream: referral traffic and lead mentions that trace back to AI tools. When someone lands on your site from an assistant, or tells you on the phone that "ChatGPT recommended you," that is the conversion evidence that separates AI search from a science project. It is early and the volume is still modest almost everywhere, so we report it plainly rather than inflating it.
Now the vanity metrics we refuse to celebrate. We ignore one-off screenshots where an assistant happened to name you once — a single lucky answer is not a ranking, because these models vary their responses and a screenshot proves nothing repeatable. We ignore raw "AI visibility scores" from tools that will not tell you what prompts they tested; an opaque number is worse than no number. And we ignore total mention counts stripped of intent, because being mentioned in an answer nobody asked while shopping is not a lead.
- What we count: repeatable citation on real buyer prompts, the accuracy of what's said, and traffic or calls that trace back to AI.
- What we ignore: lucky one-time screenshots, black-box "visibility scores," and mention counts with no buying intent behind them.
The uncomfortable truth is that AI search rewards the same foundation as everything else — a clean, well-structured site, accurate business data, and content that plainly answers real questions. Do that, measure it honestly, and you are ahead of nearly every competitor in Vienna still guessing. See how it connects to our broader AI search and SEO work.