An earlier essay here traced what happens when Google answers the question on the results page itself: the informational click disappears, and the searches that still produce patients are the ones that end in a business. That was half the story.
The other half is that many of those questions no longer start at Google at all. The owner watching their dog scratch at 11pm can now type the whole messy situation into ChatGPT — breed, symptoms, what they've tried — and get a conversation instead of ten blue links. And as of this year, there's an ad slot at the bottom of that conversation.
Every veterinary marketing vendor is about to have an opinion about this. Before anyone sells you one, it's worth understanding — in plain terms — what advertising inside ChatGPT actually is, because the mechanics decide what it can and cannot do for a practice.
How it works, plainly
Where the ads appear. Below ChatGPT's answer, clearly labeled as sponsored, visually separate from the response. The ads don't change what ChatGPT says — the answer and the ad run on separate systems, and OpenAI is explicit that advertisers can't influence the response itself.
Who sees them. Only some users: logged-in adults on the free and low-cost tiers. Paid subscribers — Plus, Pro, business, and education accounts — see no ads at all. That's an unusual shape for an ad platform: the audience is real and large, but a slice of it is walled off by design, and you can't buy your way in front of the users who pay not to see you.
How targeting works — and this is the real difference. Google Ads runs on keywords: someone types "emergency vet near me," you bid on that phrase, your ad appears against declared intent. ChatGPT ads have no keywords. Instead, advertisers write context hints — plain-language descriptions of the conversations where their service is relevant: "a pet owner describing a dog's recurring skin problems and asking whether to see a specialist." The system matches ads to conversations by meaning, not by phrase, and the hints guide it without guaranteeing anything. You describe the moment; the platform decides whether a given conversation is that moment.
What local targeting really offers. This is where a practice should slow down. Geographic controls have been maturing fast — state-level, media-market, and some ZIP-level targeting have appeared — but coverage is uneven, the tooling is beta, and nothing yet matches the radius precision a local campaign takes for granted in Google Ads. Before believing any pitch about "local ChatGPT ads," the honest step is opening the ads manager and checking whether your specific area is actually targetable. Sometimes it is. It isn't yet a promise.
How you pay. Cost-per-click or per-thousand-impressions, with daily budgets — familiar machinery. The unfamiliar part is what you get back, which is the next section.
The funnel question: who is actually in that conversation?
Picture the two pet owners who matter to a practice's marketing.
The first one's dog just ate a sock. It's 10pm, the situation is urgent, and the search is "emergency vet open now." That person is not opening a chat window to explore their options. They search, they see a name, they call. Urgent, declared, local intent still lives on the search results page — and it is still the traffic that fills an appointment book. Nothing about ChatGPT changes the sock.
The second owner's dog has been itching for four months. Two treatments haven't worked. This person has questions, context, and time — and that's exactly the conversation that has moved into ChatGPT: long, specific, exploratory. Somewhere in that thread, "should I see a veterinary dermatologist?" comes up. An ad below that conversation reaches a real future patient — but an early one. They're researching, not booking.
So the honest funnel placement: ChatGPT's ad slot sits upstream. It's an awareness position next to the researching owner, not a capture position in front of the urgent one. For most general practices, the urgent searcher is the business. For research-heavy specialties — dermatology, oncology, rehabilitation — the long-conversation owner genuinely matters, and that's where this platform is most interesting soonest.
The attribution test — the same one this site applies to everything
Here's the question every marketing dollar should survive: which patient did it produce?
Google Ads, built correctly, can answer it: the search, the click, the call, the booked appointment — a traceable line. It's the entire reason this practice-marketing model works.
ChatGPT ads, today, cannot — and the gap runs in both directions. Before you spend, there is no equivalent of Google's Keyword Planner: no way to look up what owners in your city are actually asking, no search volumes, no way to size the demand you're about to bid on. After you spend, there is no search-terms report: you will never see which conversations your ads appeared in. You describe the moment you want in a context hint, the system decides whether a conversation is that moment, and what comes back is impressions, clicks, and spend — never the conversations themselves. That's a deliberate privacy choice on OpenAI's part, and a defensible one. But it means you target demand you have to imagine, and measure with whatever your own site can see afterward. Between "an ad ran near a conversation" and "a patient booked," the line is faint. That's not a scandal; it's a young platform that hasn't built the plumbing. But it does mean that, by the standard this site applies to every marketing claim, ChatGPT ads currently fail the test not because they don't work, but because no one can yet show you whether they worked. A dollar you can't trace is a dollar on faith, and faith is what most veterinary marketing already runs on.
The verdict, calmly
For an independent practice deciding where next month's marketing dollar goes: not here yet. The searches that book patients this quarter are still on Google, the local targeting is still maturing, and the proof loop doesn't close. Moving real budget now would be trading a measurable channel for an unmeasurable one — the exact swap this site keeps arguing against.
But "not yet" is not "ignore it." Two things are worth doing today, and neither costs ad spend:
Know that the question moved. The pet-owner conversations happening in ChatGPT are the top of your funnel relocating. The practices that understand this early will make better decisions than the ones who hear about it from a vendor's pitch deck in a year.
If you must test, test small and honestly. A tightly capped budget, one research-heavy service line, context hints written as full sentences describing the exact conversation — and the discipline to judge it by tracked outcomes on your own site, not by the impression count in the dashboard. Treat it as paid research into where your future clients are, not as a patient-acquisition channel.
And the watch-list — what would have to change before this deserves real money: local targeting you can verify covers your actual draw radius; conversion measurement that connects an ad to an action you can count; and evidence, from your own site's analytics, that ChatGPT-referred visitors book. When those three appear, the calculation changes. Until then, they haven't. I've since put a date on that: four predictions about the assistant-referral era, scored publicly next August.
The test
The same closing move as always, pointed at whoever asks for your budget next.
When a vendor pitches you "AI advertising" — and they will — ask one question: "Show me how I'll know which patients came from it."
A platform that can answer gets a line in the budget. A platform that can't gets a line on the watch-list. That rule doesn't care whether the ad runs on a results page or under a conversation — and it will serve you long after this particular platform stops being new.