How we actually measure AI search — and what we refuse to count
AI search is new enough that a lot of people selling it lean on numbers that sound impressive and mean nothing. We would rather show you the measurements that map to real Bedford calls, and tell you plainly which ones we ignore on purpose. The whole point of this channel is being the name an assistant says out loud, so that is what we test — directly, by asking.
The core method is straightforward: we build the actual questions your customers ask — "who's a reliable HVAC company near Smith Mountain Lake," "best deck builder in Bedford County," "who does emergency plumbing in Forest" — and we put them to ChatGPT, Google's AI answers, and the other assistants on a set schedule. We record three things each time. Do you get named at all. Where you land when several businesses are listed. And whether the assistant describes you correctly — your services, your area, your phone number — or garbles it. That third one matters more than people expect, because a confident, wrong description sends the customer somewhere else.
We track that over time rather than as a single snapshot, because assistants vary their answers and pull from sources that update. A business that gets named in one of five test questions in month one and four of five by month three is winning, even if no traditional ranking report would show it. We also watch the sources the assistants cite when they explain a recommendation, because those citations tell us exactly which parts of your web presence the machines are actually reading — and where to strengthen it next.
Here is what we do not count, because it flatters the invoice without predicting a single call:
- Raw "AI traffic" numbers scraped from analytics — the labels are unreliable, easy to inflate, and tell you nothing about whether you were the recommendation or just a link someone skipped.
- A one-time screenshot of the assistant naming you, framed as proof — a single lucky answer is not a trend, and we will not present it as one.
- "Impressions" or reach figures borrowed from social metrics that have no bearing on whether an assistant recommends you to a buyer in Moneta.
- Vague "AI visibility scores" from tools that will not show their method, which is just a rank tracker wearing a new hat.
What connects the honest metrics is that every one of them ties back to a customer decision. Getting named, being described accurately, showing up for the specific question a lake homeowner actually types — those move business. The vanity numbers exist to make a report look busy. Because this channel is genuinely new, we are also honest about its edges: some answers are not yet measurable, assistants change without warning, and we will tell you when something is an emerging bet rather than a proven return. That honesty is the point — you should know exactly how your money is being judged. If you want to see how the assistants describe your business in Bedford today, our AI Search work starts by asking them.