How we actually measure whether AI search is working for a Franklin business
AI search — the answers people now get from ChatGPT, Google's AI Overviews, Perplexity, and Copilot instead of a page of blue links — is real, and it is quietly changing how customers around Franklin find a service before they ever click. The hard part is not getting excited about it. The hard part is measuring it honestly, because the shiny numbers are mostly noise. Here is what we watch and, just as important, what we deliberately ignore.
The metric that matters most is inclusion: when someone asks an AI assistant a real buying question — the kind a homeowner off US-258 actually types, like the best company for a specific job near Franklin — does your business get named in the answer, and is what it says about you accurate? We test these prompts on a schedule across the major assistants and log whether you appear, how you are described, and who gets named alongside you. That is the scoreboard.
The second thing we measure is the source. AI answers are pulled together from pages the model trusts, so we track which of your pages and which third-party sources — your Chamber listing, your reviews, a directory — are actually feeding those answers. If the model is citing an outdated page or a competitor's blog instead of yours, that is a concrete, fixable problem, not a mystery.
The third is referral behavior. When an AI answer does send a person to your site, they arrive already half-sold, so we watch whether that trickle of assistant-referred traffic converts at a higher rate than ordinary search traffic. It usually does, and that quality gap is the real story, not the raw count.
Now the vanity metrics we ignore. We do not chase a made-up AI visibility score with no method behind it. We do not celebrate impressions on an AI Overview when nobody named or clicked you. We do not report that AI mentioned your industry in general — that helps no one — only whether it named your business. And we do not confuse a one-time lucky answer with a durable pattern, which is exactly why we test the same prompts repeatedly rather than screenshotting a single good result and calling it a win.
The reason this discipline matters in a market like Franklin is that the field is still wide open. Most of your competitors, including the bigger ones down toward Suffolk and Norfolk, have done nothing deliberate here. The businesses that get their information clean, consistent, and quotable now are the ones AI assistants will keep naming later. But you only know you are winning if you are measuring the right things. If you want to see which prompts already name you and which name your competition, that is where our AI search work begins.