What pet owners really ask the AI: why some practices get named and others don't

Pet owners stopped searching and started asking

Three years ago a worried cat owner typed "cat vomiting cause" into Google and worked down ten blue links. Today the same person opens ChatGPT and types a whole sentence: "my cat has thrown up three times today but is eating normally, do I need to bring her in?" The answer comes back immediately, in plain language, with a course of action attached. No click, no website, no banner ad. For you, that means a good Google ranking has stopped being the same thing as being found.

The scale of this is not marginal. ChatGPT reported 900 million weekly active users in late February 2026, Google's Gemini app passed a billion monthly users in August 2026, and AI Overviews now reach over two billion people a month. A growing share of those owners reach for generative AI for a first read on a sick animal, because it is fast, patient and awake at three in the morning. That night-time hour is exactly when no practice is answering the phone and the worry is at its worst, and it is where trust gets built and specific practices get recommended.

These systems are not replacing you as the clinician. They are replacing the research path that used to end at your door. When someone asks for a vet who treats reptiles nearby, the answer either contains your practice or it doesn't, and that one sentence is now part of your new-client pipeline.

The questions that actually get asked

Questions put to an AI are longer, more specific and more emotional than the keywords people used to type. Not "dog diarrhea" but "my puppy has had diarrhea since yesterday but is drinking plenty, is that an emergency?" The shapes repeat: urgency checks ("my rabbit hasn't eaten in twelve hours, how bad is that?"), price checks ("roughly what does neutering a cat cost?") and symptom reads ("my dog is limping on the back left, what could it be?"). You already know these wordings, because they are the ones your reception team fields on the phone all day.

Then come the practice-choosing questions: "which vet near me also sees birds?", "is there an animal hospital open on Sunday around here?", "I need someone who works on horses." That is where the recommendation gets made, and the AI assembles its answer out of whatever about you is findable and structured enough to reuse.

Chronic and age-related topics come up constantly: diabetes in cats, arthritis in an older dog, dental disease, feeding a kidney patient. People research these over months rather than minutes, and they are looking for a practice that visibly does this work, not one that merely could.

Why the AI names one practice and skips another

AI systems are not picking at random, and they are not picking the prettiest site. They pull from sources they can read without guessing: your own practice site, veterinary directories, review platforms, professional writing, local mentions. What matters is how plainly your pages state who you are, which species you see, what you specialize in, and where you physically are.

A page that says "Welcome" over a contact form gives a machine nothing to work with. A page that says "we see dogs, cats, rabbits and guinea pigs; focus areas are dentistry and soft-tissue surgery; out-of-hours by arrangement" hands over the exact components an answer needs. Schema.org markup on top of that does not buy you a ranking; Google has said since 2018 that structured data is not a direct ranking factor. What it does do is make details like your hours and your service list much harder for a model to garble.

Consistency across channels does more work than most practices expect. When your name, address and hours match on your site, your Google profile and the directories, there is nothing left for a model to disambiguate. Ahrefs, looking at roughly 75,000 brands, found that how often a brand is mentioned across the web correlates with AI citation at about 0.664, roughly three times stronger than backlinks at 0.218. Third-party mentions of your practice are the best-evidenced lever there is, and contradictory details are the fastest way to get left out of an answer.

GEO for vets, in plain terms

Generative Engine Optimization, or GEO for short, is the same instinct as search engine optimization, pointed at systems that answer instead of listing. The goal is not a position on a results page; it is being the practice the answer names. And it really is a different game: an Ahrefs analysis found only about 6 to 8 percent of the URLs ChatGPT cites also sit in Google's top ten for the same query, and roughly 80 percent of them don't rank in the top hundred at all. In animal health, where a recommendation carries real weight, that gap is worth closing deliberately.

In practice, GEO means writing so that a question gets answered on the page. A section headed "when does vomiting mean my dog needs a vet today?" with a straight, clinically sound answer underneath is something a model can lift and attribute to you. Marketing filler is not. One caution straight from Google's own guidance: write for the owner reading it, not a separate track of content produced "for AI", which Google explicitly warns against as scaled content abuse.

None of this replaces local marketing. Your Google Business profile, real reviews from real clients and a site that loads and makes sense are still the floor. GEO is what keeps that floor standing when the first question goes to a chat window instead of a search box.

Make your professional authority machine-readable

Health topics get a harder look. Answer engines lean on sources with visible professional authority, so spell yours out: the names and qualifications of every vet, further training, specialist certifications, the procedures you actually perform. Letters after a name mean nothing to a system that can't connect them to a person and a practice.

Clinically solid guide content compounds this. A properly written piece on the warning signs of gastric torsion, or on what dental care in cats really involves, tells the owner and the machine at the same time that there is expertise behind the page. It also takes time to earn: Ahrefs, across 1.4 million real ChatGPT prompts, found cited pages had a median age of around 500 days. That is an argument for publishing this year rather than next.

Write medical claims the way you would say them in the consulting room, with no promises of a cure. Professional regulation requires that anyway, and it happens to be exactly what these systems treat as a credible source.

Three ways practices make themselves invisible

The most expensive mistake is vagueness. Plenty of practice sites never say outright which species they see or whether there is an out-of-hours service. A human fills that gap from context. A model doesn't guess, it omits, and you never find out you were omitted.

The second is contradiction. Different hours on your site than on your Google profile, an old practice name still hanging around after a handover, a wrong phone number in a directory. Each one hands a model a reason to prefer a source it can trust instead. Put a recurring note in the diary to check that your listings still agree with each other.

The third is treating this as somebody else's problem and trusting word of mouth to carry you. Word of mouth still works, but a younger owner asks the chat window first. If you are not in that answer, you lose those clients without a single missed call to show for it.

Three steps you can take this week

Start with the test that costs nothing: ask ChatGPT, Gemini and Perplexity for a vet in your town, and then for your particular focus. Are you in the answer at all? Are the name, address and hours right? Ten minutes of this tells you more about where you stand than any audit deck.

Second, rework your site so it answers the questions owners actually ask. A plain service list. Every species you treat, named. Your focus areas. What happens out of hours and how to reach you. Your team with their real qualifications. Then a guide section that answers the front-desk questions properly.

Third, keep the outside world consistent: the same core details everywhere, a Google profile that is genuinely maintained, reviews you have actually earned. Treat this as maintenance rather than a project, because both the engines and the way owners phrase things keep moving. Worth knowing what you can skip, too: an llms.txt file is cheap hygiene at best. John Mueller confirmed in 2025 that no Google Search system reads one, and an Ahrefs look at roughly 137,000 sites publishing one found about 97 percent got no measurable referral traffic from it.

The first contact happens before the phone rings

The first contact between a worried owner and your practice now often happens before anyone dials. It happens in a chat window, late at night, driven by a specific fear about a specific animal. Whether your practice shows up in that moment as a credible answer has become part of how new clients arrive.

The useful part is that the levers are yours. Clarity, stated credentials and details that agree with each other are not technical tricks; they are an honest description of work you already do. The systems reward the practices that bother to write it down.

GEO for vets isn't a new industry, it's ordinary practice communication carried into a world where the first question goes to a machine. Given how slowly cited pages accumulate, the practices that start now are the hard ones to catch.

A 30-day plan that fits around clinic hours

You don't need a project for this. Week one: write down the ten questions your owners genuinely ask at the desk and on the phone, word for word, from the cat that has gone off her food to whether you are open over a holiday weekend. Those exact wordings are the asset, because people type them into an AI almost verbatim.

Weeks two and three: answer each of those ten questions as its own clearly headed section on your site, with real specifics: the hours, the species, the point at which waiting stops being safe. Week four: check that your address, out-of-hours arrangements and service list read identically everywhere they appear. Details that disagree are the most common reason a model quietly picks somebody else.

One night, one question, two practices

Picture someone typing at eleven at night: "my dog has diarrhea and won't eat, do I need a vet?" The model wants a source that reads that exact situation and also gives it somewhere to send them. If you have a page that lists the warning signs and says plainly when to phone, you are the citable source.

The practice with "Welcome to our homepage" and a photo gallery is not in that answer. The gap isn't clinical skill; it is whether your competence is readable by a machine at all. That is what decides whether the worried dog owner calls you in the morning or the practice two streets over.

Where the AI gets it wrong, and what that costs you

Useful as these answers are, they are not a diagnosis, and saying so on your own site helps you rather than hurting you: sources that point toward an examination instead of promising a remote verdict read as more credible. Say plainly that a symptom list is a starting point and that anything acute needs to be seen.

Assume too that the machine will get things wrong. Columbia Journalism Review's Tow Center ran 1,600 queries across eight AI search tools and found more than 60 percent of answers misidentified their source, ranging from 37 percent for the best tool to 94 percent for the worst; ChatGPT got 134 of 200 wrong while flagging uncertainty just 15 times and never once declining to answer. These systems state incorrect things with total confidence. So when your out-of-hours cover changes, change it everywhere that day. A stale opening time is an owner standing at a locked door with a sick animal.

Treat the AI as an amplifier, not a stand-in. It will carry your expertise further than your website ever did, but only the version of it you have written down and kept current.