Generative engine optimization · Original data

One word decides whether ChatGPT cites Mayo Clinic or your business

August 27, 2026 · 6 min read

We hit ChatGPT with the same nine questions two times: first as a generic health or home-fix query, then again with a city name tossed in. The sources shifted almost entirely. Leave the city out, and the citations skew toward Mayo Clinic, Cleveland Clinic, and academic journals. Drop one in, and local businesses take over the list.

This is crucial because most owners think their invisibility issue comes down to just being a local business. It is not. It comes down to which question gets asked.

What we ran

We plugged "how do I get rid of acne scars" into GPT-4o with live web search, first as is, then again as "how do I get rid of acne scars in Miami." Same model, same day, one word added. We then ran four more city-specific questions across various trades, the kind a real customer types when they're describing a pain point instead of shopping: a migraine sufferer in Austin, someone with lower back pain in Denver, a homeowner with dropping water pressure in Chicago, hair thinning in Miami.

Without the city

The basic acne query returned eight sources. Six were institutional: Mayo Clinic, Cleveland Clinic, Harvard Health, UCLA Health, MedlinePlus, and the American Academy of Dermatology. The other two were academic, a journal article and a research repository. No clinic, no dermatology practice, no business whatsoever made the cut. If you run a med spa and someone in your area asks ChatGPT this exact question, you aren't losing to a competitor down the road. You're losing to Harvard instead.

With the city added

Tack "in Miami" onto the same question, same day, same model, and the scene flips. Five of the eight sources turned into named Miami dermatology and med spa practices: Miami Dermatology, Miami Skin Spa, Miami Skin Institute, Biscayne Dermatology, Miami Center for Dermatology. Three institutional sources stuck around, Mayo Clinic and Cleveland Clinic included, but the businesses had muscled their way into more than half the answer.

The other four questions proved the pattern holds up beyond dermatology.

Why this happens

We can't see the ranking logic behind ChatGPT's answers, but the shift is consistent enough to land on a practical conclusion. A plain informational question looks like a request for general medical or technical knowledge, so the model grabs sources built to answer exactly that: encyclopedic health sites, hospital systems, research literature. A city-anchored question looks like a request to act on that knowledge somewhere specific, so the model grabs sources that can name a place to go.

That's the trap most local businesses fall into without noticing. Owners write content answering the plain question, "what causes acne scars," "why does my back hurt," because that's the searchable, Google-friendly framing they picked up from a decade of SEO advice. But that's the exact framing where AI search routes around them to institutions with far more authority than any local practice can match. The version of the question that actually reaches them is the one with a place name attached, and that version rarely gets written as content at all.

One more wrinkle, worth knowing if you're measuring this yourself

We ran the Miami acne query through two different search backends behind the same model, OpenAI's own API and a second provider routing to the identical model. The results didn't line up. One set listed Miami Dermatology, Miami Skin Spa, Miami Skin Institute, Biscayne Dermatology, and Miami Center for Dermatology, plus three institutions. The other listed Biscayne Dermatology, Miami Skin Spa, Miami Skin and Vein, Miami Center for Plastic Surgery, and Perfectb, with no institutions at all. Only two businesses appeared on both lists.

Same question, same day, same underlying model, different search infrastructure, different answer. If you're tracking your own AI visibility, one clean result from one tool doesn't show you what every customer sees. It shows you what that tool's particular search backend surfaced that day.

What to do with this

If your content answers the general version of a question, "what causes X" or "how do I treat Y," you're competing with institutions that will usually win. Rewrite it, or write alongside it, for the version a customer actually asks when they're ready to act: the symptom plus their neighborhood, their city, sometimes their specific situation. That's the version where local sources, including yours, get a real shot at the citation.

It also means your city and service pages carry more weight for this kind of traffic than a generic blog post ever will, provided the page actually reads like an answer to the symptom question rather than a list of services. "Botox treatments in Miami" reads like a menu. "What to do about frown lines that won't relax, and where in Miami to get it checked" reads like an answer.