Before a healthcare or pharma brand in India argues about tone, it has to clear a floor that has nothing to do with tone. Some claims are prohibited outright, and no evidence rescues them. The rest are permitted only when they carry the qualifiers that make them checkable: the population, the denominator, the period, the risks, the source. Copy that skips them is not merely weak. It is unpublishable, and close to unusable to an AI answer engine.
The floor before the first draft
Before a healthcare or pharma brand in India argues about tone, it has to clear a floor that has nothing to do with tone. Some claims are prohibited outright, and no evidence rescues them. The rest are permitted only when they carry the qualifiers that make them checkable: the population, the denominator, the period, the risks, the source. Copy that skips them is not merely weak. It is unpublishable, and close to unusable to an AI answer engine.
That second half is what most teams have not connected. The disclosures a regulator requires and the attachments an answer engine needs in order to repeat you are nearly the same facts. A claim with its basis attached is one a reviewer can approve and a model can quote. A bare superlative fails review and gets hedged away in the answer.
Why are medicines and clinical services under different law?
Because one governs a product and the other governs a service, and each starts from a different assumption. Claims about medicines, devices and diagnostics sit under the drugs regime administered through CDSCO, alongside the Drugs and Magic Remedies (Objectionable Advertisements) Act 1954. The opening position there is that a therapeutic claim is allowed when it is true, substantiated and within the approved indication.
Clinical services sit somewhere else. A hospital, clinic, imaging centre or fertility unit advertises a service delivered by registered practitioners, and the NMC Code of Ethics begins from the position that practitioners may not solicit patients, directly or indirectly, at all. That is not a stricter version of the drugs rule. It is a different question. Not is this claim substantiated, but is this communication solicitation.
Which is why these are two rulebooks, not one with a wider scope line. Three prohibitions on the clinical side are criminal rather than regulatory, and none has an equivalent in a medicines rulebook. A diagnostics brand screened only against a drugs pack would pass copy that is an offence to publish.
The claims prohibited outright, not merely unevidenced
Most compliance work is about evidence: attach the study, name the population, state the limits. A smaller set of claims sits outside that logic. No source you can attach makes the sentence publishable, because the prohibition attaches to the act of advertising rather than to whether the claim is true.
The consequence is a review order. Evidence questions close by attaching a source, so they belong inside the drafting loop. Prohibition questions never close, so they belong before it. A campaign built on one of the sentences above does not need a stronger citation. It needs a different idea.
Cure, prevention or treatment claims for conditions scheduled under the Drugs and Magic Remedies Act, including cancer, diabetes, heart disease, HIV, kidney and liver disorders, infertility and obesity. A trial result does not convert these into permitted public copy.
Advertising prescription-only medicines (Schedule H, H1 or X) to the public. That content may address registered medical practitioners only.
Prenatal sex determination, or any service implying it, a criminal offence under the PCPNDT Act 1994.
The supply of, or demand for, human organs, an offence under the Transplantation of Human Organs and Tissues Act, 1994.
Commercial surrogacy and the recruitment of surrogates, prohibited under the Surrogacy (Regulation) Act 2021.
Guaranteed clinical outcomes and solicitation by superiority: "guaranteed cure", "assured recovery", "100% success", "best hospital", "top surgeon".
Absolute safety or efficacy language for any medicine, device or treatment: no side effects, completely safe, works for everyone.
Audience decides legality more often than tone does
The rule most marketing teams do not know exists is that the same sentence can be lawful addressed to a prescriber and unlawful addressed to a patient. Dosage, course and regimen for a prescription medicine belong in prescriber communication, not on a public page.
So the audience question is answered at the page level, not the campaign level: a gated practitioner portal and a public blog can carry the same facts and stand in different positions.
AI answers press on exactly this seam. An answer engine has no audience gate. It reads what it can fetch and re-states it to whoever asked, so a patient asking a plain question can receive a passage lifted from material written for a clinician. If a page is reachable, treat its audience as the public, and gate prescriber-only material for real.
Why is a success rate without its denominator unusable?
Because a percentage on its own does not describe anything. A fertility success rate quoted per cycle and the same clinic's rate quoted per patient are different numbers about different things. Change the age band or the period and the figure moves again. Two clinics can publish the same figure and mean quite different things, so a patient cannot compare them and neither clinic can defend its own.
So the requirement is not to hold evidence but to publish the basis next to the number: per cycle or per patient, the age band, the period, and the source. That is the shape a claim has to be in before anyone, a regulator, a patient or a machine, can check it.
The same failure repeats wherever a fact is separated from its qualifier. A package price without what it excludes: implants, consumables, room category, complications, follow-up. An accreditation named without the accrediting body, the entity accredited and the current validity, when accreditation is granted per facility and expires. A screening result presented as diagnostic when it only indicates. The number or the badge is real, and the sentence is still not checkable.
What does an answer engine do with a claim it cannot verify?
It hedges the claim, attributes it to somebody else, or leaves it out. An answer engine builds its reply from sources, repeating the statements it has reason to trust. A claim arriving with its population, denominator, period and source gives the model something to stand on and a reason to cite you. A bare superlative gives it nothing, so the model softens the sentence into a generality or routes around it to a source that did the work.
Health questions get the most conservative treatment of any category. Assistants qualify medical answers heavily and steer readers towards a qualified practitioner, which is right, and which also means an unsupported marketing claim rarely survives into the answer intact.
Read that in the other direction and it is an opportunity. The pages quoted in health answers state a specific claim, name its basis, and are honest about its limits. That is the same page compliance wanted. The disclosure you treated as a tax on the copy is what makes the copy citable.
The disclosures that travel with the claim
The must-include rules are less famous than the prohibitions, and they are where most drafts actually fail. They share one logic. A benefit never travels alone.
Ansyra screens drafts against these packs live in the editor, with the rulebook chosen by what the brand actually does, and separates what blocks a publish from what only needs a source. It then measures whether AI answers about the brand are accurate, so a corrected claim can be shown to be gone. One honest limit: a passing screen establishes that these rules matched nothing, and nothing more. It is a starter rulebook, not legal advice, and your own legal and compliance teams remain the approvers.
State the recognised side effects, contraindications or risks alongside a therapeutic benefit, not on a separate page.
Name the generic (INN) alongside the brand whenever a specific medicine is discussed.
Carry a consult-a-qualified-practitioner line on public health content, and never position content as a substitute for diagnosis or prescription.
State that a telemedicine service is not for emergencies, and direct urgent cases to in-person care, per the Telemedicine Practice Guidelines 2020.
Name the qualification and registration of any practitioner referenced, without implying a specialisation that is not registered.
Say what a diagnostic result does and does not confirm on its own.
State how reports, images and records are handled, consistent with the DPDP Act.
Attach a source to every efficacy, safety, comparative, outcome, success-rate and accreditation claim, and route the approval-gated ones to a reviewer first.