Six months into an AI search visibility campaign, most practices cannot answer the simplest question about the work: is it working?
Google Analytics shows organic traffic. Rank-tracking tools show keyword positions. Neither report answers whether ChatGPT is recommending the practice to prospective patients, whether the recommendation is flattering, or whether the answer is changing over time. This is the measurement gap in GEO (Generative Engine Optimization), and it is the reason a lot of marketing dollars going into AI search work are being spent without honest confirmation that they are landing.
The Measurement Problem
Traditional SEO measurement was built for a world where visibility ended at a search engine results page. Rank-tracking tools report positions on a defined keyword list. Analytics tools report the click from that result to the practice’s website. Between the search and the click there is a straight line, and every dashboard in the industry was built to track it.AI search does not work that way. When a patient asks ChatGPT for the best cosmetic dentist in a particular neighborhood, the platform generates an answer. The patient reads it. The click, if it happens, is optional. Whatever visibility occurred, occurred inside a conversation no analytics tool has direct access to. The measurement layer that mattered for two decades cannot see the new one.
That leaves practices with a genuine problem. GEO work is being commissioned, invoiced, and reported on, and the traditional dashboard does not confirm the outcome the practice actually cares about.
What a Practice Can Track
There are four categories of signal available today. None of them is perfect. All of them are better than nothing, and together they build a defensible picture.
Direct prompt testing. Someone on the team runs a fixed set of patient-realistic prompts through ChatGPT, Perplexity, Gemini, and Claude on a regular cadence, typically monthly. The prompts are the ones a real patient would type: who is a good cosmetic dentist in a specific neighborhood, which dentist handles clear aligners nearby, recommend a family dentist near a landmark. The test records whether the practice is named, how it is described, and whether the description flatters or damages the practice’s positioning. Results compare month over month. This is manual, imperfect, and by a wide margin the most direct signal available.
AI Overview visibility on Google. Google Search Console does not report whether a practice appeared in an AI Overview, but several third-party tools estimate it against a defined query set. A practice should not treat these estimates as precise counts. They are a directional gauge on tracked queries, useful for spotting movement rather than for auditing a single week’s performance. Because AI recommendations differ from traditional Google rankings, a practice can lose AI Overview presence on a query where its blue-link ranking is unchanged, or gain AI Overview presence on a query where its ranking never moved.
Referral traffic patterns from AI platforms. When a patient does click through from ChatGPT or Perplexity, the referrer often shows up in analytics. Absolute numbers are usually small. Rising referral traffic from AI platforms is a positive indicator. Flat traffic is harder to interpret, since most patients who see the practice in an AI answer never click through, and the absence of clicks does not mean the absence of visibility.
New-patient intake signals. The most honest measurement question a practice can ask is downstream. Are new patients arriving who cite AI as part of their research process? A single well-designed intake question, followed by a probing follow-up when the answer is vague, captures more useful information than most rank-tracking reports. The data is soft. The signal is real, and it correlates with the outcome the practice cares about.
What a Practice Cannot Track Yet
Honesty matters here more than confidence.
A practice cannot count precisely how many times ChatGPT recommended it last month across every conversation on the platform. That data lives inside the AI companies, and none of them publish it. A practice cannot fully separate the AI-driven portion of its organic growth from the traditional-SEO-driven portion, because the two share underlying signals (dental schema markup, authoritative citations, clean local presence) and reinforce each other in ways that do not decompose cleanly. A practice cannot lock in a per-visibility-improvement dollar figure, because the attribution chain from an AI answer to a booked appointment is too long and too noisy for that kind of math.
The measurement environment for AI search resembles local SEO measurement around 2010: directional signals, improving month by month, useful without being precise.
A Better Question to Ask
The question “how do we measure AI search performance” is real, and the answer is imperfect. A more useful question sits underneath it: what would a practice need to see over the next six months to conclude that its GEO investment is working?
Reasonable answers include a clear lift in direct prompt-test results across the top AI platforms, a rising share of new patients citing AI in their research process at intake, steady growth in referral traffic from AI platforms, and stable or improving performance on the AI Overview queries where the practice competes. None of these signals is definitive on its own. A practice that sees movement across most of the four categories is winning in AI search. A practice that sees no movement in any of the four probably is not, regardless of what the traditional dashboard reports.
The measurement problem in AI search is real, and no marketing partner should pretend otherwise. What a practice should expect from a partner is a monthly reporting layer that covers all four signal categories, honest framing of what each signal can and cannot say about the work, and a willingness to revisit the measurement approach as the AI platforms release new tools.
Bring those criteria to a discovery call and see what an AI search visibility report should look like for a specific practice.
