How we tell whether AI search is actually working for you
AI search is new enough that plenty of people sell it on vibes and screenshots. We'd rather tell you exactly how we measure whether it's working for a Front Royal business, and which numbers we ignore because they look impressive and mean nothing.
The metric that matters is citation share. We take the real questions your customers ask — a good HVAC company in Front Royal, a canoe outfitter near the Shenandoah River, where to eat before Skyline Drive — and run them across ChatGPT, Google's AI Overviews, and the other tools on a repeating schedule. The scorecard is simple: for each prompt, are you named, and where in the answer. We track that over time. If you were cited in two of ten relevant prompts in month one and seven of ten by month three, the work is landing. That's the real dashboard.
Underneath that we watch the inputs the AI models actually pull from, because those are the levers. Is your business described the same way everywhere the model can read? Are your reviews and mentions consistent and findable? Does the page an outfitter or contractor points to actually say, in plain language, what you do and where — Front Royal, the Shenandoah River, Warren County — or does it make the model guess? When one of those inputs improves and your citation share climbs a few weeks later, we know the mechanism is working, not just the outcome, and we know which lever to pull again.
Here's what we deliberately ignore. Raw AI traffic as a headline number is close to a vanity metric today — most AI tools don't reliably pass a clean referral, so a big or small figure there tells you very little. Screenshots of a single lucky answer are worse; ask the same question twice and the model may phrase it differently, so one flattering result proves nothing. And total impressions or mentions divorced from your actual customer questions are noise — being named in an answer nobody in the Shenandoah Valley is asking doesn't fill your calendar.
- We report: citation share across your real prompts, position within the answer, and whether that share is trending up month over month.
- We ignore: one-off screenshots, raw AI referral counts, and impressions on questions your customers don't ask.
The tie-breaker is always the phone and the booking form. We ask new customers how they found you and watch for the AI recommended you showing up, because that's the outcome citation share is a proxy for. Done right, AI search is measured like everything else we do — against real questions, tracked over time, and judged by whether it moves customers, not dashboards.