Ranking on Google but Not on ChatGPT

We hear a version of this from new clients often enough that it's worth writing down properly: "we rank on the first page for our main keyword, so why did I ask ChatGPT the same thing and it recommended someone else?" It's not a bug, and it's not bad luck. Google's ranking algorithm and a language model's answer generation are doing genuinely different jobs, and a page can satisfy one without coming anywhere near the other.

Two different questions, not one

Google's core job, even now, is still mostly retrieval: given a query, return the most relevant pages, ranked by a few hundred signals accumulated over two decades. A language model answering a question is doing something closer to synthesis. It's not handing back a list of links, it's generating a single answer, and it has to decide, in the moment, which sources it trusts enough to draw from. Those are different tasks with different winning conditions.

A page built around solid keyword placement, decent backlinks, and reasonable content depth can genuinely earn a page-one ranking. None of that guarantees the page reads as trustworthy or extractable to a model deciding what to cite. We've seen practices with a five-year-old, well-linked site rank comfortably on Google while a newer competitor with a more structurally clean site gets cited more often by AI tools, despite fewer backlinks and less domain age.

A concrete example

Picture two dermatology practices in the same city, both ranking on page one for "acne scar treatment." Practice A's page is well-written but structured as a long single block of text under one heading, no schema, no FAQ section, testimonials scattered through the body without attribution. Practice B's page has clear H2s for each treatment type, a physician byline with credentials stated in the first paragraph, MedicalClinic and Physician schema in the code, and a short FAQ block answering "how many sessions does laser resurfacing take" directly. Both pages might rank similarly on Google. Ask an AI model the same question, and Practice B is far more likely to get named, because there's simply more for the model to confidently extract and repeat.

What actually seems to matter for citation

Structure matters more than most practices expect. A model parsing a page for facts does better with clean heading hierarchy, explicit schema (MedicalClinic, Physician, FAQPage), and content that states things plainly rather than burying an answer in the fourth paragraph after two paragraphs of brand introduction. Pages that read fine to a human but are structurally flat under the hood, common on a lot of older WordPress themes, are harder for a model to extract clean facts from.

Consistency matters too, more than it seems like it should. If your name, address, or credentials are stated slightly differently across your website, your Google Business Profile, and a couple of directory listings, a model has no clean way to resolve which version is authoritative. It doesn't investigate the discrepancy. It just leans toward whatever version shows up most consistently, and that's not always your own site.

And attribution carries real weight. Content clearly written and reviewed by a named, credentialed physician reads as more trustworthy, both to Google's E-E-A-T evaluation and to a model weighing whether a claim is safe to repeat. Anonymous content, even when it's accurate, tends to get treated more cautiously by both systems.

How to test this yourself, step by step

You don't need any special tools to check where you currently stand. Open ChatGPT, Perplexity, and Gemini in three tabs. In each one, type the exact question a patient would type, not a keyword phrase, an actual question: "who's a good [your specialty] in [your city] for [a specific concern]?" Note three things for each answer: whether your practice is mentioned at all, whether the facts stated about it are accurate if it is, and which practices get named instead if it isn't. Repeat this monthly. Citations shift as content ages and as competitors do their own structural work, so a one-time check tells you where you stand today, not where you'll stand in three months.

The misconception worth clearing up

The most common misunderstanding we run into is that GEO means writing more content. It usually doesn't. Most practices we audit already have enough content, the problem is structural: no schema, inconsistent business details across the web, no clear physician attribution. Our GEO service covers exactly this. Fixing those three things on existing pages typically moves the needle faster than writing new pages ever could, and it's a technical project with a defined scope, not an open-ended content commitment.

Where to start if you're auditing this yourself

Begin with the three factors that correlate most consistently with citation: schema markup, business-information consistency, and named physician attribution. Our LLM SEO service covers the technical implementation of the first, and our GEO service covers ongoing citation testing once the foundation is in place. Business-information consistency in particular means something slightly different depending on which regulatory bodies apply, our UAE page covers how DOH and DHA licensing details factor into this specifically.

Quick answers