Most AI visibility checks test a question the AI is designed not to answer

At some point in the past year, you have probably run the test yourself. You opened Google, typed "best ENT near me," or the version for your specialty, and looked at what came back. Maybe there was no AI answer at all, and you privately concluded this whole AI-visibility conversation is overblown. Or maybe a vendor ran a free scan built on that same query and sent you a score with some red on it.

Either way, you were looking at a question where the answer means almost nothing. Not because AI does not matter for how patients find doctors, but because that is the question Google's AI has stopped answering.

The answer that was taken away on purpose

BrightEdge, a search firm that tracks how often Google shows an AI answer across query types, followed provider-intent searches, queries like "dermatologist near me," through the transition. In December 2023, when Google's AI answers were still an opt-in preview, essentially all of these queries triggered one. A year later, 14 percent. By December 2025, zero.¹

That was not decay. A coverage line that goes from everything to nothing in two years reads as a decision, not drift. For the moment of actually choosing a doctor, Google walked its AI out of the room and handed the job back to the map, the profiles, and the reviews.

The same tracking shows the opposite on everything upstream. Symptom, condition, and treatment questions trigger AI answers 93 to 100 percent of the time.¹ "Why is my ear ringing." "Sinus pressure or migraine." "Does a deviated septum always need surgery." Ask those, and the AI answers at length.

So a blank on the near-me query tells you nothing about your practice. A score built on it, comforting or alarming, is measuring the same dead spot: the one question where the engine now stands down.

The choosing happens upstream

Set referrals and insurance-directory lookups aside. The self-directed patient's journey rarely starts at "best ENT near me." It starts weeks earlier, with a symptom typed at 11pm. Then: is this serious. Then: what are the options. Then: is this worth seeing a specialist about. Only at the end, and sometimes never, comes the query with a specialty in it.

On Google, the early and middle rungs get AI answers almost every time, and those answers are assembled from sources: clinic pages, health sites, community threads. Whether your practice's material is among those sources shapes whether you exist in the conversation a patient has been having for days or weeks before they ever look at a map.

And on the conversational engines, the ones patients type whole paragraphs into, the entire ladder happens in one thread, and names do come back. A patient who has asked four symptom questions often gets a provider suggestion in the same conversation. The practices that surface there did not win a near-me query. They were present in the material the engine had already been reading.

What this means for every check you are offered, including ours

Full disclosure: we give away a free check that asks one engine one patient-style question about your practice. We think it is a useful screening question, and we try to be plain about its scope. But a screening question is not a workup, ours included. One question, on one engine, on one day, is a data point, not a diagnosis.

So hold every visibility check, free or paid, to a skeptic's standard. It should tell you which questions it asked, and why those. Which engines it asked, since they read different sources and often disagree. What a blank means on each surface, because on Google's AI a blank on a provider query is the designed behavior, not a finding. And the date, because answers drift.

A check that cannot answer those four things is an opinion with a number attached.

The fifteen-minute version you can run yourself

You do not need a vendor for the honest test. Three questions, in a patient's words, not yours:

One symptom question your patients actually bring in. One treatment question, with the procedure named the way a patient says it. One "is this worth seeing a specialist about" question.

Run them on Google and on one conversational engine. Note three things: whether an AI answer appears, whether any practice gets named, and what kind of source the answer leans on when one is. Fifteen minutes, and you will know more about your actual position than the near-me test could ever tell you.

If a practice gets named and it is not you, resist the conclusion that you have failed. Absence is the norm for independent practices right now, and the reasons are usually specific and fixable. But that is a different article.

The near-me test feels right because it mirrors twenty years of search habit. The engines moved the conversation upstream. The useful measurement moved with it.