This issue is about the discovery layer, which is the one part of the sector where change is observable in real time rather than pending in a consultation. It is also the part where the sector's commentary is least disciplined about the difference between an observation and a sales pitch.
Discovery and regulation intersect at one point and it is worth naming: the advertising rules constrain what can be said in the exact places where demand is highest.
A patient searching a toxin brand name is expressing high intent, and the advertiser cannot lawfully meet that query with a promotional page for the medicine. A patient searching a symptom is expressing lower intent, and the advertiser can meet that query freely with explanatory content. The rules therefore push supply towards the symptom side of the market, which is also where the discovery shift is pushing it.
ObservedThe prohibition on advertising prescription only medicines to the public is statutory and mirrored in the CAP Code.AnalysisTwo independent forces are pointing the same direction. That is unusual and it makes the strategic conclusion unusually robust.You do not have to choose between the compliant content strategy and the effective one. In this category, for the moment, they are the same strategy.
The mechanical difference matters. Keyword retrieval matches strings and rewards pages that repeat the string. Embedding based retrieval matches meaning and rewards pages that cover a concept comprehensively, in the vocabulary the asker used, with enough context to be extractable as an answer.
The practical consequences: long, specific, well structured explanatory pages outperform thin procedure pages; question shaped headings are retrievable in a way that marketing headlines are not; and precision about jurisdiction, definitions and limits becomes a ranking asset rather than a legal chore.
ObservedGenerative answer surfaces cite sources visibly, which means the set of pages being drawn on is partly inspectable rather than entirely opaque.AnalysisThe unit of competition is shifting from the page to the passage. A page can be irrelevant overall and still contain the one paragraph that answers the question, and that paragraph is what gets used.SpeculationIf passage level citation continues to grow, site architecture matters less than paragraph level clarity. That would be a genuine reversal of a decade of SEO practice, and we would not bet a business on it yet.Write paragraphs that survive being extracted alone. Self contained, specific, and stating the qualification inside the sentence rather than three paragraphs later.
Institutional sources are the natural beneficiaries of retrieval based discovery, because they are what a system optimising for reliability would prefer to cite.
The regulators, the registers and the clinical bodies in this sector publish material that is authoritative, stable and poorly optimised. That combination is a citation opportunity for anyone willing to explain the institutional position accurately and link back to it, and it is a warning to anyone whose content contradicts it.
AnalysisA page that contradicts the regulator is not merely wrong, it is competing against the source the system trusts most in the category.Alignment with the primary source is now a discovery strategy. That is a strange sentence to write and it appears to be true.
Patients arrive with a diagnosis they got from a machine, and they arrive with the vocabulary of the machine's answer.
This is not new in medicine. What is new is the fluency and the confidence. A generative answer reads as authoritative and rarely signals its own uncertainty, so patients arrive attached to a framing rather than to a question. Consultation practice built around correcting misconceptions from forums is not calibrated for misconceptions that arrive pre argued.
AnalysisThe change is in the confidence of the misconception, not in its existence. Correcting a confident framing takes longer and creates more friction than correcting a vague one.SpeculationPractitioners who publish accurate explanatory content may find they are correcting less, because the source the patient consulted was theirs. That is a plausible mechanism and we have no data on it.The cheapest consultation is the one where the patient arrives already correctly informed. That makes explanatory publishing an operational investment, not a marketing one.
"We are optimised for AI search." There is no such certification and no published standard to be optimised against. The claim is a proxy for having written clear content, which is worth doing, described in a way that suggests a technique that does not exist.
Traffic and visibility figures for AI answer surfaces. Treat every one you are shown with suspicion. Attribution from generative answer surfaces is partial and inconsistent, and confident percentage claims about it are being made well ahead of the measurement.
ObservedReferral attribution from generative answer surfaces is incomplete by construction, because a cited answer frequently satisfies the query without a click.If someone shows you a number about AI visibility, ask how it was measured and what it excludes. In our experience the answer is usually a sample of prompts run by hand, which is a legitimate method and a very different thing from a measurement.
The economic consequence of the shift is a change in the shape of the acquisition funnel, not necessarily in its size.
If an answer engine satisfies the informational stage of the journey, the visits a clinic receives are fewer and later. Fewer visits at higher intent is not a worse outcome, but it breaks every metric calibrated on volume, and it makes content look like it is failing at exactly the point it is working.
AnalysisThe measurement problem arrives before the commercial problem. Businesses that cut informational content because sessions fell may be cutting the thing that produced the remaining enquiries.SpeculationEnquiry quality is the metric that survives this transition. Session volume is the metric that does not.Change what you measure before the numbers force you to. Enquiries, consultations booked and conversion from consultation are all still measurable. Sessions are becoming a proxy for nothing in particular.
The direction is towards fewer, later, better informed arrivals, and towards a category in which the entity being cited matters more than the page being ranked.
That favours publications and clinics with a coherent, consistent, verifiable identity: a clear name, a consistent description, accurate structured data, and a body of work that says the same thing across every surface. It disfavours businesses whose online presence is a set of unrelated pages that happen to share a domain.
AnalysisEntity coherence is cheap to build and slow to fake, which is exactly the property a retrieval system would want to reward.SpeculationWe expect the sector to respond to this with a wave of services promising entity optimisation. Most of what is required is unglamorous consistency, and it does not need a vendor.Say the same true thing about yourself everywhere, and make it checkable. That is most of the work, and nobody can sell it to you.