The same query, two different outputs

Type "best dentist in Boulder" into Google in 2015 and you get ten blue links, a map pack of three to five practices, and probably some paid ads. You scroll, you compare, you click a few, you decide. The search engine's job was to surface options. Your job was to evaluate them.

Type the same query into ChatGPT, Perplexity, or a Google AI Overview today and the experience is different in kind, not just in style. You get a paragraph. It names a practice, maybe two, gives a brief reason, and stops. There is no page ten to scroll to. There is no "see more results" for a curious patient to click.

This is not a redesign of the search results page. It is a different task being performed. A traditional search engine ranks; a generative engine answers. Ranking is a sorting problem — put the most relevant pages near the top and let the human do the rest. Answering is a judgment problem — the AI system has to decide, on the patient's behalf, which practice is worth naming.

That distinction is the entire reason GEO (generative engine optimization) exists as a discipline separate from traditional SEO. A practice can be excellent at the sorting problem — solid keywords, decent backlinks, a fast-loading site — and still lose the judgment problem, because the judgment problem is being solved with different inputs.

What changes when an answer replaces a list

The practical consequences of this shift are easy to underestimate until you sit with the math.

The floor for visibility rises sharply. On a traditional results page, a practice ranking seventh out of twelve local competitors is still visible — a patient willing to scroll a bit will find it. In an AI answer that names one to three practices, ranking fourth is functionally identical to ranking twentieth. There is no partial credit for being a strong runner-up.

The criteria shift from ranking signals to trust signals. A search engine's ranking algorithm cares about things like keyword relevance, backlink profiles, and page experience metrics. An AI system building a recommendation cares about a different, overlapping but distinct set of things: can it verify who you are, what you offer, whether you are still in business, and whether other patients vouch for you — quickly, and from more than one source. Some of the same signals matter to both (a well-built website helps either way), but the AI system's bar for confidence is different from a ranking algorithm's bar for relevance.

The output is opinionated, not exhaustive. A results page implicitly says "here are your options, you decide." An AI answer implicitly says "here is who I recommend." Patients treat these differently. A patient scrolling ten results applies their own judgment to each one. A patient reading a single AI-generated recommendation is far more likely to simply act on it — call the number, book the appointment — because the AI system has already done the evaluating for them.

Why "best" in the AI answer does not mean objectively best

It is worth being direct about something practice owners often assume incorrectly: the practice named in an AI answer is not necessarily the best dentist in Boulder by any clinical or reputational measure. It is the practice whose online presence gave the AI system enough verifiable, corroborated information to name with confidence.

This distinction matters because it means AI visibility is a solvable, buildable problem rather than a reflection of who has practiced longest or has the fanciest office. In our research analyzing how AI systems answer local business queries across the Front Range — more than 70 businesses studied across seven batches of realistic search queries — we consistently found smaller, newer practices outranking larger, more established competitors in AI-generated answers. The determining factor was not size, tenure, or even review volume alone. It was how completely and consistently the practice's facts were represented across the sources an AI system draws on: Google Business Profile, the practice's own website, and review platforms.

A well-regarded practice with twenty years in Boulder and a loyal patient base can be entirely absent from AI answers if its Google Business Profile is thin, its website lacks structured data, and its reviews are old and sparse. Meanwhile a newer practice that has built a complete, verifiable, consistently-represented online presence can capture the recommendation instead. This is not a flaw in the system from a marketing standpoint — it is an opportunity. The gap between "objectively good practice" and "AI-recommended practice" is exactly the gap that a deliberate GEO strategy closes.

How an AI system actually builds the answer

Understanding roughly how these systems construct a recommendation helps explain why some practices get named and others do not.

When a patient asks an AI assistant for a dentist recommendation, the system is not consulting a single ranked database. It draws on a combination of sources — its own training data, live web search results when the platform supports it, structured data on business websites, Google Business Profile information (directly or indirectly, depending on the platform), and review content from Google and other platforms. It then synthesizes those inputs into a short, confident answer.

The practices that get named tend to share a few characteristics that make this synthesis easier:

  • Their basic facts (name, address, phone, hours, services) are consistent across every source the AI system might check, leaving no contradictions to resolve.
  • Their website uses structured data that states facts explicitly rather than requiring the AI system to infer them from marketing copy.
  • Their reviews are recent, detailed, and specific enough to serve as evidence the AI system can point to.
  • Their content directly answers the kind of question being asked, rather than requiring the AI system to guess whether a generic "General Dentistry" page is relevant to a specific query like "pediatric dentist good with anxious kids."

Practices missing several of these elements are not necessarily excluded on purpose. They are simply harder for the AI system to verify quickly, and an AI system optimizing for a confident, defensible answer will gravitate toward the practice that requires the least inferential leap.

What this means for a Boulder-area practice

If your practice has not been showing up when you or your staff test AI search queries, the honest diagnosis is rarely "our patients don't like us" or "we're not well known." It is much more often a structural, fixable gap: incomplete profile data, missing schema markup, thin or stale reviews, or content that never answers the specific questions patients are actually asking an AI system.

The corollary is genuinely good news for practices willing to do the work. Because AI visibility is largely a function of structure and completeness rather than size or budget, a well-organized effort can move a practice from invisible to recommended in a matter of weeks for the foundational pieces, with reviews and content depth compounding over a longer horizon.

More on this topic

If you want to know whether your practice is one of the names AI systems are currently recommending in your market, our free AI Visibility Audit at novasapienlabs.com/audit tests your visibility across real patient queries and shows you exactly where the gaps are.