A woman in Fort Collins types a sentence into her phone at 9 p.m. because her dog won't stop limping. "Good vet near me that's gentle with anxious dogs." She doesn't open ten blue links. She reads the answer the assistant gives her, which names two clinics and describes each in a line or two. One of those lines is about you. Or it isn't, and it's about the practice four blocks away.
That sentence is the whole introduction now. Not your website. Not your lobby. A compressed verdict, written by a model, delivered before the pet owner has looked at a single photo of your waiting room.
The review nobody wrote
Here's what's strange about it. No one wrote that sentence. Not you, not a reviewer, not a journalist. It's a synthesis. The assistant read across whatever it could find about your clinic and produced a characterization that stands in for all of it. Call it reputation compression: hundreds of separate human opinions, squeezed into one confident-sounding claim a stranger treats as fact.
People trust it because it sounds like a friend who happens to know every vet in town. It reads as neutral. It isn't neutral. It's a reading, and the reading can be generous, flat, or quietly damaging, and you had no say in which.
This is already happening. Ask any of the seven major assistants — ChatGPT, Claude, Gemini, Perplexity, Google AI Overviews, Grok, Copilot — for a good vet in a mid-sized city and you'll get named clinics with little verdicts attached. Not a page of ten results to scroll. A short, named shortlist, which means the difference between being in it and being out of it isn't a lower rank. It's near-total absence. The pet owner never learns you exist.
What the model is actually doing to your reviews
A pet owner reading your reviews sees a stream of individual voices. Someone's cat came through surgery fine. Someone waited too long in the lobby. Someone cried at the front desk and a tech sat with them. A human weights all that intuitively and forms a feeling.
The model does something colder. It looks for what repeats. A single glowing review barely moves it. A single angry one barely moves it. What moves it is a pattern that shows up again and again, and especially a pattern that shows up the same way in more than one place. Agreement across independent sources is what lets a model state something as fact instead of hedging. When the reviews on one platform and the description on another and the way you present yourself on your own site all point the same direction, the model grows confident, and confidence is what earns you the clean, warm sentence. When those sources contradict each other — your site says "24-hour emergency care" and the reviews are full of people who couldn't reach anyone after six — the model hedges. And a hedging model reaches for someone it can describe more cleanly. You don't get a bad sentence. You get skipped in favor of a clinic the model felt sure about.
That's the mechanism underneath the whole thing. Not star count. Consistency of signal. The assistant is trying to say something it won't get wrong, and it rewards the clinics that make that easy.
Take Google Business Profile, for instance
Of all the places a model looks, your Google Business Profile does unusual work, so it's worth understanding why rather than just hearing that it matters.
Picture two clinics in Tucson, both genuinely good, both staffed by people who care. The first clinic's profile is alive. The hours are right, including the Saturday half-day. The services listed match what the front desk actually does. The reviews are recent and the owner replies to them in a steady human voice — thanking the happy ones, addressing the rough ones without getting defensive. When a client writes "they got my senior lab in same-day when she stopped eating," the words sit there in public, in the client's own language, describing a specific thing that happened.
The second clinic is just as good in real life. But its profile hasn't been touched in two years. The hours are stale. Half the services aren't listed. The reviews are strong but old, and nobody replies to them, so the page reads like a building with the lights off.
Now the model comes through. It's not grading these clinics on kindness, which it can't measure. It's looking for corroboration — information that appears, confirmed, in a place it trusts. The first clinic hands it exactly that. The reviewer said "same-day for a sick senior dog" in plain words, the profile confirms the clinic is open and operating, the owner's reply confirms the clinic is present and engaged, and the whole thing agrees with itself. The model can now write "well-reviewed for responsive care with older pets" and feel safe doing it. That sentence came from your reviewers. The profile is what let the model trust it enough to repeat.
The second clinic gives the model nothing recent to stand on. Great care, no current evidence of it. So the model does what models do under uncertainty: it hedges, or it moves on to a clinic it can describe without risk. The care was identical. The compressed verdict wasn't, because one clinic made agreement easy to find and the other left the model guessing.
Notice what this isn't. It isn't a form to fill out in order. The profile matters because of what it does — it gives an assistant a trustworthy place to confirm what your clients are already saying, so their words survive the compression into that one sentence. Understand that and the specific tasks stop feeling like busywork and start feeling like the point.
This isn't a theory about the future
We watch this across verticals, and veterinary medicine isn't exempt. The RankCommander AI Visibility Index has now evaluated tens of thousands of real AI answers across local service categories, and still counting, and the same thing keeps surfacing: two businesses of genuinely comparable quality, one named warmly by the assistants and one invisible, with the difference tracing back to how legible and consistent their public signal was. You can see the vertical benchmarks and the disclosed sample sizes at /ai-visibility-index — real counts, not a number we made up to sound impressive.
The uncomfortable part for a clinic owner is how little the strength of your actual medicine protects you here. You can run the best practice in the county and be the one the model hedges on, because the model never set foot in your building. It read the description. The described clinic is the one that gets the call.
The client you're already losing
Think about who that limping-dog owner in Fort Collins becomes. A new puppy next spring. Dental cleanings for a decade. An emergency on a holiday weekend where she pays whatever it costs because it's her dog. A friend she tells at the dog park. The lifetime value of one new household walking in the door is enormous, and you're not losing a line on an analytics dashboard when the assistant names someone else. You're losing her, and everyone she would have referred, to a competitor who did nothing more impressive than be easier for a model to describe.
And the competitor is specific. There's a clinic in your city that is, right now, the name an assistant offers when a worried owner asks. Maybe it's the newer practice that's organized about its online presence. Maybe it's the one you've always considered a notch below yours on actual medicine. It doesn't matter. In the sentence the pet owner reads at 9 p.m., that clinic is the recommendation and you're not in the paragraph. Every month that stays true is a month of first-time clients forming a bond with someone else's front desk. Loyalty, once it starts, is hard to pry loose. The window where you could have been her clinic closes quietly, and you never see the appointments you didn't get.
That's the loss. Not slow erosion. Clients who were deciding tonight, decided, and you weren't an option they were given.
What to do about it
The hardest part of all this is that it's invisible from inside your practice. Your schedule looks fine. The phones ring. Nothing announces that the assistants have been quietly handing your new-client conversations to the clinic across town. You'd have to ask all seven of them, in your own city, over and over, and read every sentence carefully enough to catch what's wrong, and know what to do about the gap when you find it. No practice owner has time to run that by hand.
RankCommander does it for you. We scan what ChatGPT, Claude, Gemini, Perplexity, Google AI Overviews, Grok, and Copilot are actually saying about your clinic right now, show you the real sentence each one gives a pet owner, tell you which competitors are getting named in your place, and track it as it moves instead of leaving you guessing. You've spent years earning the trust of the families who walk in. Don't let a stale description hand the next family to someone else. Run your scan and find out, today, what the assistants are telling the owner whose dog is limping tonight.