AI Visibility Is a Retrieval Problem, Not a Ranking Promise

There is no reliable “number one ranking” in ChatGPT.

Ask the same healthcare question with a different location, wording, context, or date and the answer can change. Some responses search the web and cite sources. Some answer from other available context. Some decline to recommend a provider at all.

That makes a familiar SEO promise—“we will get your practice recommended first”—especially suspect. A practice does not control the answer system. It can control whether its public information is accessible, specific, consistent, and supported well enough to be used with confidence.

AI visibility is best treated as a retrieval and representation problem, not a guaranteed rank.

Begin with the questions that matter

“Does ChatGPT know our brand?” is too broad to guide work. Build a small set of questions based on real patient decisions and factual business needs.

Some questions are categorical: Which practices in this area offer a specific service? Some are evaluative: What should I look for in a provider? Some are factual: Where is the practice, which clinicians work there, and how does consultation work? Some are educational and should not be forced into a brand recommendation.

Define the audience, geography, service, and decision stage for every test. Avoid testing only “best” queries designed to produce a winner. In healthcare, a responsible answer may emphasize credentials, evaluation, risk, or the limits of a generalized recommendation.

Make public information retrievable

OpenAI’s current publisher FAQ says public sites can appear in ChatGPT search and advises publishers not to block OAI-SearchBot if they want content considered for summaries and snippets. It also explains how publishers can track referral traffic from ChatGPT search using the UTM parameter OpenAI adds to referral URLs.

That is a concrete technical check. It is not a promise of inclusion. Crawl access allows discovery; it does not establish that a page is the best source for a particular answer.

Apply the same discipline to ordinary search. Pages need to be accessible, indexable where intended, internally coherent, and clear about their subject. Google’s current guidance emphasizes helpful, reliable, people-first content, not content produced primarily to attract search visits.

Publish evidence that resolves uncertainty

Many healthcare pages make claims that are difficult for any reader—human or machine—to verify. “Advanced care,” “leading provider,” and “personalized treatment” reveal very little.

A useful service page identifies the actual clinicians, location, scope, relevant credentials, process, candidacy boundaries, and sources for factual medical claims. A location page states which services are available there. A provider biography distinguishes licensure, training, certification, memberships, and experience accurately rather than blending them into a prestige cloud.

Structured data can restate those facts in machine-readable form, but it should match visible content. Schema does not convert a weak website claim into evidence or force an answer engine to cite the page.

Third-party sources matter when they independently verify something important: a licensing board for license status, a certifying body for certification, a professional organization for membership, or reputable coverage for a public event. Directory volume by itself is not authority, and inconsistent listings create ambiguity.

A worked test: mention, citation, accuracy, preference

Suppose a multi-location orthopedic group tests: “Where can I get a sports injury evaluated near Northside this week?”

One answer names the practice but sends the user to the wrong location. A second cites the practice’s service page without naming it in the summary. A third names two competitors and cites a directory. Calling all three “rankings” loses the useful information.

Score separate dimensions. Was the practice mentioned? Was an owned page cited? Were location, service, clinicians, and access facts accurate? Was the practice presented as one option, preferred, or excluded? Which sources shaped the answer?

The first priority is correcting the wrong location across owned and authoritative sources. Trying to increase mention share before fixing the factual error would scale confusion.

Measure a changing system honestly

Create a repeatable prompt set and a recording method:

• Save the exact question, date, product, model or mode when shown, and relevant location or account context.
• Capture the complete answer and cited URLs, not just whether the brand name appeared.
• Classify factual accuracy, citation, mention, recommendation posture, and competitors named.
• Repeat on a defined cadence and label the sample size and limitations.
• Track referral traffic and downstream outcomes separately from answer observations.

Do not present a handful of manual prompts as market share. Personalization, experimentation, source availability, and product changes can affect results. The measurement is a controlled observation of selected questions, useful for diagnosis and trend direction.

A single missing mention is not proof that the practice is invisible everywhere, either.

Improve the source, not the prompt theater

When a test exposes a gap, repair the underlying information. Consolidate conflicting location facts. Add a substantive page for a real service. Clarify a provider credential. Cite primary medical or regulatory sources. Earn legitimate third-party coverage by doing something worth covering.

Avoid manufacturing reviews, templated directory pages, synthetic “expert” quotes, or dozens of near-duplicate answers. Those tactics create more text but less confidence.

AI answer systems add another interface between a practice and the public. The durable strategy is not to reverse-engineer a secret recommendation formula. It is to make the practice easy to understand, difficult to misstate, and worth citing—then verify whether that is happening.

---

If you want to measure AI visibility without turning a few prompts into a vanity score, request a growth plan. Decabrand can define a question set and trace the sources behind the answers.