We Asked ChatGPT and Perplexity to Pick Healthcare Marketing Agencies. They Barely Agreed.
Nick
·6 min read
Table of contents
TL;DR: In a small pilot, we asked ChatGPT and Perplexity the same three questions about hospital marketing agencies in the US, UK and UAE. Together they named 51 agency slots, 45 different agencies, and only 6 appeared on both tools. Perplexity leaned on a single listicle for most of its US answer, while ChatGPT leaned on agencies' own websites. If a patient or hospital buyer asks AI who to hire, which tool they use changes who they hear about.
Picture a hospital marketing director who skips Google and asks an AI tool for a shortlist of agencies. Then, to double check, she asks a second tool. The two shortlists look like they came from different industries.
We wanted to know if that happens in practice. So we ran a small, honest test.
The pilot: three questions, two tools, one afternoon
On October 22, 2026, we asked ChatGPT and Perplexity the same four questions, one run each. Three asked for the best healthcare marketing agencies for hospitals and clinics in the US, the UK, and Dubai and the UAE. The fourth asked what a new patient costs a medical practice. ChatGPT ran in temporary chat with memory off, and we asked it to search the web and cite sources.
Read this as a pilot, not a statistic. It is one run per question, on free accounts, from a browser located in India. AI answers change from run to run, and we could not test Gemini. We kept the full log, and we are sharing the headline numbers rather than pretending this is a representative sample.
Almost nobody appeared on both lists
Across the three agency questions, the two tools filled 51 agency slots. Those slots held 45 distinct agencies. Only six agencies were named by both tools.
The overlap by market:
Market | ChatGPT named | Perplexity named | Named by both |
|---|---|---|---|
United States | 6 | 8 | 3 |
United Kingdom | 10 | 7 | 2 |
UAE | 10 | 10 | 1 |
The UAE was the starkest. Ten agencies on one list, ten on the other, and a single agency in common. A buyer who asked only one tool would never know the other list existed.
This fits a pattern in larger studies. Semrush's AI Visibility Index found that ChatGPT and Google AI Mode mention many of the same brands but share only about 30% of their sources. Different sources, different answers.
Perplexity leaned on one article, ChatGPT leaned on the agencies themselves
The more interesting finding was where the answers came from.
For the US question, Perplexity cited ten sources, and one list article from a single publisher supported six of its eight agency rows and most of its buying advice. Only a few other sources appeared. In plain terms, one article shaped most of the answer a buyer would read.
ChatGPT worked differently. Its US and UK answers mostly pointed to each agency's own website and case studies, with a small number of third party comparison pages and directories. For the UK and UAE it also pulled in regulators, citing government advertising guidance and a Dubai Health Authority circular on medical advertising.
Both approaches have a catch. If AI leans on agency websites, you win by publishing clear, specific case studies on your own site. If AI leans on a handful of list articles, you win by being included in them, which is a link and PR problem, not just a content problem.
What the big studies say, and why they disagree
Larger studies point in different directions on this exact question. A study of 6.8 million AI citations from Yext found that 86% of citations come from sources brands manage themselves, such as websites, listings and reviews. A separate 8,400-prompt UK tracker found brand-owned sites earned only 12% of citations, with most answers citing third parties instead.
Both can be true. Citation share depends heavily on the type of query, the industry, whether listings count as owned, and which tools are tested. Our pilot sits between them: ChatGPT looked closer to the Yext picture and Perplexity looked closer to the UK tracker. The honest conclusion is that no single "AI visibility" number tells you where you stand. You have to check the specific questions your buyers ask, on the tools they use.
Even the cost answers split
We also asked both tools what a new patient costs a medical practice. They overlapped on one source, a benchmark report from First Page Sage, out of nine distinct sources cited. ChatGPT suggested roughly $150 to $350 per new patient as a starting benchmark, while Perplexity landed on about $300 to $400. Both reported a specialty range running from about $155 in pediatrics to about $610 in cosmetic or plastic surgery.
One more detail worth noting: ChatGPT added a table of Indian lead costs without being asked, apparently because of where the browser was located. AI answers are not the same for every searcher, which makes any single "we ranked first" claim harder to trust. If you are building content around numbers like these, a practical guide to healthcare content marketing covers how to present them accurately.
We were not in the answers either
For the record, our own agency did not appear in any of the eight responses. We are not hiding that, because it is the point. Ranking well in Google does not automatically carry over to AI answers, and being absent from a shortlist is invisible unless you go looking for it. Most agencies, and most clinics, have never run this check on themselves. The first time you do, expect to find a few surprises, including competitors you had not heard of and sources you had never tried to earn coverage from.
What this means if you sell healthcare services
Four practical takeaways, none of which require a bigger budget:
Test the questions your buyers actually ask. Run them on at least two AI tools and record which agencies and sources appear. Do it monthly, because answers move.
Publish proof on your own site. Case studies with named outcomes, clear service pages and specific specialties gave agencies a reliable path into ChatGPT's answers.
Get into the list articles. If one article drives most of a tool's answer, being listed there matters. That means earned coverage and white hat link building, not just more blog posts.
Make your regulatory knowledge visible. In the UK and UAE answers, advertising rules came up repeatedly. Agencies that explain local rules clearly give AI something worth citing.
If your clinic or hospital needs help turning this into a plan, our team works on healthcare content marketing for hospitals and clinics and local SEO for doctors.
Two tools, two shortlists, and a buyer who only sees one
The real finding is not that AI is unreliable. It is that AI is inconsistent in a way that rewards attention. Two questions asked in the same afternoon gave two different markets, and the gap was widest where competition is thinnest.
That makes this a good moment to measure rather than guess. A monthly check of the questions your buyers ask costs an hour and tells you which sources are shaping your reputation before a client ever calls.
Pick five questions your buyers ask, run them on two AI tools this week, and write down who shows up and what they cite.
Frequently asked questions
Does being cited by AI mean people will click through to my website?
Not necessarily. A source can be named in an answer without a link being followed, so track branded searches and enquiries alongside citations.
Should I optimize for ChatGPT or Perplexity first?
Start with whichever your buyers use, which you can learn by asking new patients or clients how they found you. If you do not know, the practical overlap is small enough that testing both is worth the hour.
How often should I check what AI says about my business?
Monthly is a sensible starting point, and again after any large website change or press coverage. AI answers can change from one run to the next, so repeat important questions more than once.
Can a small clinic show up in AI answers at all?
Yes, particularly for narrow questions such as one procedure in one city. Smaller providers tend to do better when their pages answer a specific patient question directly.
Is it safe to rely on AI answers when hiring an agency?
No. Both tools told us to verify case study claims independently, and agency reported results are not audited. Treat an AI shortlist as a starting list, then ask for references.
