Week 3 · The 6 Compliance Regimes and How to Write Clean Copy

Part of the ICG ChatGPT Ads Training Series · Week 3 of 5 · By Rohit Gupta

What you'll learn this week

By the end of this week you should be able to look at a piece of ad copy and know, without asking anyone, which of the six regimes it needs to clear and why. That is a more useful skill than memorising every clause of every code, because in practice you will not be citing sections — you will be catching a claim before it goes to your compliance reviewer and wastes a review cycle.

The six regimes we cover are the National Medical Commission guidelines on practitioner advertising, the ASCI Code for advertising claims generally, the Digital Personal Data Protection Act 2023 for anything involving a lead form, the Assisted Reproductive Technology (Regulation) Act for fertility-specific claims, the Uniform Code for Pharmaceutical Marketing Practices for anything touching medicines or devices, and the AYUSH advertising rules for Ayurveda, Yoga, Unani, Siddha and Homeopathy providers.

Before this week, you should already know how conversational sponsored responses are assembled and served (Week 1) and how the intent-first auction decides which response wins a given conversation turn (Week 2). If either of those is fuzzy, revisit them first — compliance review makes a lot more sense once you understand that your copy is being pulled into a live conversation, not sitting static on a page waiting to be read.

This week does not turn you into a lawyer, and it should not replace your actual legal or compliance sign-off process. What it gives you is a first-pass filter, so that the copy your team sends for review has already had the obvious problems removed. That alone cuts review cycles down substantially, because most rejected copy fails for the same handful of predictable reasons — reasons this week walks through one at a time.

Core concept 1 — NMC and ASCI: the two regimes that touch almost everything

Start here because these two apply to nearly every healthcare ad you will ever write, regardless of specialty. The National Medical Commission's guidelines on practitioner advertising exist to stop registered doctors from turning their professional standing into a marketing asset in ways that could mislead patients or create unfair competitive pressure among practitioners. In practice, this means individual doctor names generally stay out of sponsored ad copy. The clinic or hospital brand carries the promotion; the doctor's credentials get referenced factually on the landing page, where a patient has chosen to seek that detail out, rather than pushed at them inside a conversational response they did not ask for.

This distinction — promotional surface versus informational surface — is the single most useful mental model for NMC compliance. A sponsored response saying "Dr. Sharma is Delhi's top cardiologist" fails immediately: it names a practitioner, it makes a superiority claim, and it does both inside a promotional context. The same information, reframed as "Cardiology care at [Hospital Name], with a specialist team experienced in interventional procedures," carries the same underlying fact — an experienced cardiologist is on staff — without triggering either problem.

ASCI's Code for Self-Regulation runs alongside NMC and covers advertising claims more broadly, not just healthcare. Three categories of claim get flagged most often in our reviews. Unsubstantiated superiority claims — "best," "top-rated," "number one" — need documented evidence to survive scrutiny, and most healthcare providers do not have the kind of independently verified ranking data that would support them, so these words get removed rather than defended. Guaranteed-outcome language — "guaranteed results," "100% success," "pain-free recovery" — is close to an automatic rejection in medical advertising, because no clinical outcome can honestly be guaranteed at the individual level. And comparative claims that imply inferiority in unnamed competitors — "unlike other clinics," "better than typical treatment" — fail even without naming anyone specific, because the comparison itself is the problem, not the missing name.

The fix for all three categories is the same move: replace comparison and superlative language with specific, factual, checkable detail. "Best-in-class equipment" becomes "3T MRI installed in 2025." "Highest success rates" becomes a specific, sourced statistic if one exists, or gets dropped entirely if it does not. This is not a euphemism exercise — it is a discipline of only claiming what you can actually stand behind, which happens to also be what both regimes require.

One nuance worth flagging: ASCI review looks at the ad in context, including the landing page it points to. Clean sponsored copy that lands on a page making the exact claims you removed from the ad does not actually solve the problem — it just moves it one click downstream. Compliance review for a ChatGPT Ads campaign has to cover the full path, not just the visible ad text, which is the point covered in more depth under the worked example later in this page.

Core concept 2 — DPDP Act 2023: consent has to be built into the funnel, not bolted on

The Digital Personal Data Protection Act 2023 changes what a lead form is allowed to do the moment it collects a name, phone number, or any health-related detail — which, for a healthcare lead form, is essentially always. The Act requires that consent be specific, informed, and freely given, which in practical copywriting terms breaks down into three requirements your form and its surrounding copy need to satisfy.

First, the purpose of collection has to be stated in plain language before the person submits the form — not buried in a linked privacy policy they are unlikely to open. "We'll use this number to call you about appointment availability" is compliant framing; a bare phone-number field with no context is not, even if a privacy policy link sits somewhere on the page. Second, consent needs to be itemised rather than bundled — a single checkbox covering "marketing communications, data sharing with partners, and appointment reminders" all at once does not meet the specificity bar, because a person cannot meaningfully consent to three different uses through one gesture. Third, there has to be a working, reasonably easy mechanism for someone to withdraw consent later — a phone number or email that actually reaches a person who can act on the request, not a dead link.

ChatGPT Ads adds a wrinkle here that a Search or Meta ad does not have in the same way: the sponsored response itself is part of a conversation, and the platform's own data-handling sits upstream of anything your landing page does. Your compliance responsibility starts at the point where your own systems take over — the moment someone clicks through and reaches your lead form. That is the boundary this week's checklist audits, and it is also the boundary most teams get wrong, because they assume platform-level data handling covers their own form's obligations. It does not.

In practice, the DPDP fix that resolves most issues we find is a short consent block placed directly above the submit button, written in the same plain language as the rest of the page, itemising exactly what each piece of information will be used for and naming a real contact point for withdrawal. It reads like three sentences, not a legal paragraph, and it is worth testing — teams who worry it will hurt conversion typically find it does not, because specific, transparent language builds more trust with a healthcare lead than vague reassurance does.

The cost of getting this wrong is not abstract. A DPDP complaint against a healthcare lead form carries real penalty exposure, and unlike an ASCI or NMC issue, which typically results in a takedown request, a DPDP violation is a statutory compliance failure. This is the regime worth being most conservative about, because the downside is structurally different from the others.

Core concept 3 — the three specialty-specific regimes: ART, UCPMP, and AYUSH

The last three regimes do not apply to every campaign — they apply based on what you are advertising, and missing that trigger is the most common gap our compliance passes catch. The Assisted Reproductive Technology (Regulation) Act governs fertility clinic advertising specifically, and it is stricter than general healthcare rules on several fronts. Success-rate claims are heavily restricted — a fertility clinic advertising a specific IVF success percentage needs that figure to be substantiated and appropriately contextualised, because success rates vary enormously by age, diagnosis, and protocol, and an unqualified headline number is close to definitionally misleading. Any language that could be read as inducement toward a particular reproductive choice, and anything adjacent to sex-selection, sits outside what any compliant campaign can touch, full stop.

The Uniform Code for Pharmaceutical Marketing Practices becomes relevant the moment ad copy references a specific medicine, device, or manufacturer-linked offer — which happens more often than teams expect, particularly for diagnostic chains promoting a named imaging technology or a hospital promoting a branded surgical device. UCPMP restricts how pharmaceutical and device promotion can be framed, and the safest practical rule is to describe capability rather than brand: "advanced robotic-assisted surgery available" clears more easily than naming the specific manufacturer's system, unless there is a documented reason the brand name itself needs to appear.

AYUSH advertising rules — covering Ayurveda, Yoga, Unani, Siddha and Homeopathy — are, in several respects, the strictest of the six. Disease-cure claims get scrutinised particularly tightly here because of a long history of misleading advertising in traditional-medicine marketing specifically, and regulators have responded with correspondingly tighter enforcement. Copy for an AYUSH provider needs its own claims pass beyond the general ASCI check — words like "cure," "reverse," or "permanent relief" attached to any AYUSH treatment are close to automatic rejections, and disclaimers required for specific condition categories need to be present, not just implied.

The practical takeaway across all three: before writing a single line of ad copy, identify which specialty-specific regime applies, and run that check first, before the general ASCI and NMC pass. Teams that reverse this order — writing generically compliant copy and then discovering a specialty trigger late — end up rewriting from scratch more often than teams that flag the trigger at the brief stage.

Worked example — an anonymised ICG engagement pattern

One recurring pattern from our engagements illustrates how these regimes stack in a single piece of copy. A multi-specialty hospital group client — anonymised here, as is our practice for engagement examples — came to us with a ChatGPT Ads draft written by an internal team before compliance review, intended to promote a cardiac care package.

This single paragraph failed on four separate counts at once, which is a fairly typical density of issues for a first draft written without a compliance pass in mind. It named an individual practitioner in a promotional context — an NMC issue. It used "top" and "guaranteeing" and "best possible outcomes" — three separate ASCI-flaggable superlative and guarantee claims in one sentence. It made an implied-inferiority comparison against unnamed competing hospitals — another ASCI issue. And its lead form collected name and phone number with zero stated purpose and zero consent language — a DPDP gap.

The rewrite our compliance pass produced kept the underlying value proposition — an experienced cardiac team, a structured care package, an easy next step — while resolving all four issues:

Notice what changed and what did not. The specific, factual detail — years of combined experience, the structure of the care package — stayed and, if anything, got more concrete, because concrete facts are what survive an ASCI review while vague superlatives do not. The doctor's name came out; the hospital brand carries the promotion instead. The comparison against unnamed competitors disappeared entirely, because it was adding rhetorical heat without adding information. And the lead form gained one plain-language sentence that states purpose and offers withdrawal — which, in our experience of running this rewrite pattern across cardiac, oncology, and fertility campaigns for hospital clients, does not measurably hurt click-through into the form and in several cases modestly improved lead quality, because it filters toward genuinely interested prospects.

This case also illustrates why compliance review has to check the full path, not just the ad snippet. The original landing page for this campaign repeated the "top cardiologist" and "guaranteed outcomes" language in its hero section — meaning that even if the sponsored response had been clean on its own, the click-through destination would have reintroduced the same violations. Our review scope for every ChatGPT Ads engagement covers ad copy and landing page as one unit for exactly this reason.

Exercise for you to do this week

Pull the last three pieces of ad copy your team has written or is planning to write for any paid channel — ChatGPT Ads or otherwise, since this exercise works regardless of platform. For each one, run through the following pass and note what you find.

1. Does the copy name an individual doctor or practitioner? If yes, could the same information be carried by the clinic or hospital brand instead?
2. Circle every superlative, comparative, or guarantee word — "best," "top," "guaranteed," "unlike," "number one," "cure," "proven." For each one, ask whether you have documented, checkable evidence behind it. If not, either find the evidence or replace the word with a specific fact.
3. Does the specialty trigger ART, UCPMP, or AYUSH review? If the campaign touches fertility, a named medicine or device, or an AYUSH modality, flag it for the specialty-specific pass before general review.
4. Open the landing page the ad points to. Does it repeat any claim you just flagged in the ad copy? If so, the fix needs to happen on both surfaces, not just the one you started with.
5. Read the lead form's consent language, if any exists. Does it state a purpose for each field collected, and does it offer a real way to withdraw consent? If it is a bare form with no surrounding text, that is the single most common gap we find, and it is usually the fastest one to fix.

Keep a written note of what you found and what you changed for each piece — not because anyone is grading it, but because building the habit of documenting a compliance pass is exactly what the ICG pre-launch checklist below formalises into a repeatable process. Teams that write this down consistently catch problems earlier and spend far less time on review cycles later.

Common failure modes to avoid

A few patterns show up often enough across engagements that they are worth naming directly, so you recognise them before they cost you a review cycle.

• Treating compliance as a copy-only exercise. The most common failure is cleaning the ad text while leaving the landing page untouched, as the worked example above shows. Review the full path every time.
• Assuming a claim is safe because a competitor uses similar language. Other providers running non-compliant ads are not evidence that the approach is safe — they are simply a compliance risk you have not been asked to inherit.
• Bundling consent into one generic checkbox. "I agree to be contacted" covering marketing, data sharing, and appointment reminders all at once does not meet DPDP's itemised-consent requirement, even if it feels efficient.
• Missing the specialty trigger. Fertility, pharma-adjacent, and AYUSH campaigns each need an additional pass beyond general ASCI and NMC review — skipping it is the single most frequent reason a specialty campaign gets sent back.
• Softening claims with hedge words instead of removing them. "Among the best" or "one of the top" does not resolve a superiority claim — it just adds a qualifier around the same underlying problem, and most reviewers will flag it just as quickly.