A prospective patient in Scottsdale opens ChatGPT and types "best pediatric dentist near me." Three practices come back with warm, specific recommendations. A fourth practice — older, better-reviewed, arguably better at the actual dentistry — isn't mentioned at all. The dentist who runs it has no idea this conversation happened. He'll never see it in any report he currently reads. And the family that asked the question already booked with one of the three.
This is the part of the shift that catches people off guard. You can't manage what you can't see, and most local professionals cannot see this at all. Their rank-tracking tools still show them holding position. Their analytics platforms still show traffic. Meanwhile a growing share of buying decisions is being routed through assistants that never generate a click, never show up in a referral report, and quietly make a recommendation that either includes you or doesn't. Measuring that is a different discipline than measuring search rankings. Most businesses haven't started.
Why the old dashboards go quiet here
Traditional search told you where you stood. Position four for a keyword, position one for another, a trend line you could read at a glance. The whole model assumed a ranked list of blue links and a human doing the choosing.
An AI assistant collapses that. It reads across many sources, forms a view, and answers in a sentence or two that names a handful of businesses — often just one. There's no page two. There's no "you're close, keep pushing." You're in the answer or you're absent, and absence looks identical to never having existed. Google Search Console can show you impressions on your own domain, but it can't tell you that Gemini recommended the practice across town this morning when someone asked which orthodontist takes their insurance. That conversation leaves no trace in any tool built for the click era.
So the first job of measurement is admitting the old instruments have gone dark, and building new ones that watch the actual behavior: what the assistants say when a real customer asks a real question.
The metric everything else hangs off
Start with the one number that matters more than any dashboard flourish: how often you get named. Take the questions your customers actually ask at the moment they're ready to act, and ask them across ChatGPT, Claude, Gemini, Perplexity, Google AI Overviews, Grok, and Copilot. Count how often your name appears. That proportion — your citation rate — is the spine of the whole practice, because it maps directly to reality. Every query where you don't appear is a customer who heard a recommendation that wasn't you.
What makes this genuinely hard, and why a casual check misleads, is variance. Run the same query twice and the wording changes. Sometimes the roster of names changes. A dentist who tries this once, sees himself, and relaxes has measured nothing. He caught a single roll of the dice. The signal lives in repetition under controlled conditions: the same question set, the same platforms, the same time of day, run again next month so the two are actually comparable. Change the questions between runs and you've compared apples to a different fruit. This is the difference between a measurement and a screenshot, and it's the reason most DIY attempts stall out. People underestimate how much discipline consistency requires.
Around that spine, a few other readings turn a raw score into a diagnosis. Whether you show up more on some assistants than others tells you the shape of the problem — visibility is rarely uniform across all seven. Whether the AI describes you accurately matters, because being named with the wrong specialty or a closed location is its own kind of invisible. And who occupies the answers you're missing from tells you exactly what you're up against.
What "who's in your gaps" actually reveals
That last reading deserves a closer look, because it's where measurement stops feeling like a report and starts feeling personal. When you run your query set and you're absent, someone else is present. Not a directory. A named competitor. The practice three neighborhoods over. The firm you beat at the bar association lunch. Their name is the answer a real person just received to a question you'd have loved to answer yourself.
Understanding why they're there and you aren't means understanding what an assistant is doing when it forms a recommendation. It isn't taking anyone's word for it. A model doesn't recommend a business because the business claims to be good. It recommends the business whose story holds together across independent sources — the practice that shows up described the same way in more than one place that doesn't depend on the practice itself. Agreement across sources reads as trustworthy. Contradiction reads as risk, and a model that senses risk hedges, and a hedging model reaches for a safer name instead of yours.
Take Healthgrades as a worked example of how this plays out for medical practices. It's not that an assistant has been told Healthgrades is authoritative in some hardcoded way. It's that Healthgrades is one of those independent places where a physician's specialty, location, credentials, and patient experience get stated in structured form that a model can read and cross-check. When what a doctor's own site says lines up with what shows up there — same specialty, same address, same name spelled the same way — the model sees corroboration. Two independent sources telling the same story. Confidence goes up. When the site says one thing and that external record says another, or the external record is thin or stale, the model has contradictory inputs and no clean way to resolve them, so it grows cautious about that name and moves on to a doctor whose profile doesn't create that friction. The competitor sitting in your gap usually isn't beating you at medicine. They're beating you at coherence — their story agrees with itself everywhere a model looks, and yours doesn't yet.
That's the mechanism. It's not a secret and it's not magic. It's why "we have great reviews" doesn't automatically translate into being recommended, and why two practices with similar reputations can land on opposite sides of the same AI answer.
Turning readings into a report you'll actually act on
Measurement that sits in a folder changes nothing. The point of tracking citation rate, cross-platform breakdown, accuracy, and who's displacing you is to build a feedback loop with a rhythm you can maintain.
Monthly is that rhythm. Weekly tempts you to react to noise — a name drops out one week, panic, then it's back the next and you've wasted a fortnight chasing a coin flip. Quarterly is worse in the other direction. A full season is long enough for a competitor to become the default answer in your market while you're not looking, and every week they hold that spot is a stack of prospects who asked, got their name, and booked. Never yours to lose in the traditional sense, because you never knew they were shopping. That's the cruelty of this channel. The losses are invisible until you measure them.
A month gives the variance room to average out while keeping the loop tight enough to respond. You run the standardized set, you compare it honestly to last month's under identical conditions, and you watch the direction of travel. Rising means what you're doing is compounding. Flat means the market moved and you didn't. Falling means someone's actively taking ground. If you want to go deeper on the underlying number before you start tracking it, our breakdown of the AI visibility score covers how the components fit together, and the GEO guide covers the optimization side once you know where you stand.
What "good" even means here
There's no universal passing grade, and anyone who quotes you one is selling a number stripped of context. A cosmetic dentist in a metro packed with competitors and a probate attorney in a rural county are not measured on the same scale. Being named in most of your relevant queries is strong anywhere. Almost never surfacing is a warning anywhere. Between those poles, "good" is defined by your vertical and your market, not a round number someone put on a slide.
This is exactly why benchmarking against real data beats chasing a threshold. RankCommander's AI Visibility Index publishes per-vertical benchmarks built from thousands of evaluated AI answers, so a dentist can see where dentists actually land and a lawyer can see the same for law, rather than guessing against a generic target. Knowing that the typical practice in your category sits where it sits reframes your own number from an abstract score into a competitive position. And it makes the trend line the thing you watch — not "did I hit 60," but "am I closing the distance on the names ahead of me every month." If you're already comfortable with tracking search performance more broadly, this is the same instinct pointed at a channel most of your competitors haven't measured yet.
Here's what separates the practices winning this from the ones losing it: they know their number, and they watch it move. Not the mechanics, not the tactics — the plain fact of having a real, current reading while their competitors are still flying on rank reports from the click era.
See where you actually stand
Right now, today, an assistant is answering a question a customer in your city just asked. It's naming somebody. The only question that matters is whether it named you or the practice down the road who figured this out first. You've spent years building a reputation that deserves to be the answer — and none of that reputation helps if the systems people now ask can't see it clearly enough to say your name. Every month you don't measure this is a month a competitor can quietly become the default recommendation in your market, one invisible booking at a time, and by the time it shows up in your revenue it's already a habit patients and clients have formed around someone else. Don't guess. Run your free AI visibility scan and see exactly what the assistants say when someone asks for what you do — before the answer hardens into a name that isn't yours.