How we measure whether AI search is actually working
AI search is new enough that a lot of people selling it lean on numbers that sound impressive and mean nothing. Since your customers here are the engineers and analysts who build these systems, they'll see through vanity metrics faster than most — so it's worth being precise about what I track and what I deliberately ignore.
The number that matters most is inclusion: when someone asks ChatGPT, Google's AI Overview, Perplexity, or Gemini a real buying question in your category — "who's a reliable roofer in Ashburn," "best pediatric dentist near Broadlands" — does your business show up in the answer, and how are you described when it does? I run those queries on a schedule across the assistants your customers actually use, log whether you appear, note whether you're named accurately, and watch that presence build over time. That's the scoreboard. Everything else is in service of moving it.
Underneath inclusion, I track the inputs that drive it, because AI recommendations are downstream of things you can influence. That means whether your site clearly and machine-readably states what you do, where you do it, and why you're credible; how consistently and accurately you're described across the third-party sources these models read; and the volume and recency of the reviews and mentions that signal you're a real, trusted operator. When inclusion moves, I can point to which of these inputs moved first — so it's cause and effect, not a coincidence you're paying for.
The last real metric is the one that pays the bills: qualified contacts you can trace to AI-driven discovery. When a caller says "ChatGPT recommended you" or a lead lands after an AI Overview surfaced your name, that's the signal that inclusion is turning into revenue. It's harder to attribute cleanly than a Google click, and I won't pretend otherwise, but it's the difference between being cited and being chosen — and it's the number worth chasing.
Here's what I don't report, because it flatters the invoice without telling you anything. Raw "AI traffic" counts that lump in bots and one-off curiosity. A screenshot of your business in a single answer to a single query on a single day, when the same question tomorrow returns someone else. "Optimization scores" from tools that grade your page against a checklist no AI system actually uses. And any promise of a fixed rank or guaranteed placement — these systems don't rank in a stable list, they synthesize fresh answers, and anyone guaranteeing a spot is guessing. What I'll show you instead is honest and durable: are you in the answers your Ashburn buyers are actually getting, are you described the way you'd want, and is it bringing you work. If those three are climbing, it's working. If they're flat, I'll tell you and we'll change the approach.