Working notes, not thought leadership.
Everything here carries the data behind it, including the findings that do not flatter us. Where we got something wrong, the correction is published under its own heading rather than edited in quietly.
Why “healthcare marketing agency” in India is a map query, not a search query
Ten agencies splitting about 1,900 visits a month, decided by street address. The full dataset.
The medical advertising code that actually applies in India (it is not the one you think)
The 2023 regulations were put in abeyance within weeks. The 2002 code is what governs you.
The one thing a diagnostic centre may never advertise
Section 22 of the PCPNDT Act, and how it is breached by implication rather than intent.
Fifty patient queries, four AI systems: the protocol, and a correction.
A pre-registered study of how often answer engines name an actual clinician.
Where AI Overviews actually appear in health search
Which health queries trigger an AI Overview, and which still return ten blue links.
Three files that decide whether ChatGPT can see your hospital
robots.txt, the sitemap, and llms.txt. Most hospital sites get at least one wrong.
Where “47% of patients use AI to find a doctor” came from
We traced a statistic everyone quotes back to its origin. It does not survive the trip.
Author schema: the markup your consultant’s profile page is missing
The fields that tell a machine your consultant is a real, registered, identifiable person.
E-E-A-T and YMYL for clinical websites: what the framework actually asks for
Stripped of the folklore, and applied to a hospital site rather than a blog.
Your robots.txt is probably blocking the AI that should be citing you
Eight crawlers, what each one feeds, and the one line that turns you invisible.
Core Web Vitals on a hospital CMS: where the four seconds go
Measured on real hospital sites. The culprit is almost never what the agency says.
Doctor profiling done properly: video, long-form and the schema that connects them
One forty-minute conversation, and everything you can responsibly build from it.
Media selection for healthcare: which platform, for whom, and why most hospitals get it backwards
Short form, long form, and the platforms that cannot legally do what you want them to.
What frightened people actually read at 2am
From building a second-opinion intake flow. It is not what the marketing textbooks predict.
Marketing spend versus waiting-list length: the diagnostic nobody runs first
How to find out whether marketing is your problem before you spend anything on it.
What running a PET-CT second-read service taught us about patient trust
We built the thing, not just the marketing for it. Here is what it changed.
Why we run the Second Read platform on profit-share instead of retainer
If the enquiries do not arrive, we are not paid. What that does to the work.
The work we turn down, and why
Four kinds of engagement we decline, stated plainly so you can self-select out.
Disambiguating a brand name three other companies already owned
We chose the name before we checked the index. Published rather than quietly fixed.
Every piece states its method and the date its data was pulled. Search figures come from Ahrefs and are twelve-month averages carrying normal estimation error — treat them as directional. Legal and regulatory claims are sourced to the regulator’s own text, never to a summary of it. Anything touching a clinical decision is written by a non-clinician and says so; it is marketing analysis, not medical guidance.
We would rather be corrected than quoted.
Every table we publish is reproducible on any Ahrefs seat in under an hour, and the keyword lists are stated in full. If you run one of these and get a different answer, we would like to hear about it.