How a pharma brand ran Scale-tier UCPMP-clean ChatGPT Ads for a Rx oncology-support product with disease-awareness-only messaging
The situation
Picture an Indian pharmaceutical company with a prescription-only oncology-support product — a supportive-care therapy administered alongside primary chemotherapy to manage a specific class of treatment-related complication, prescribed exclusively by treating oncologists and dispensed through hospital and specialty pharmacies rather than open retail. The brand had a well-established medical-affairs presence — conference symposia, peer-reviewed publication support, oncologist-facing detailing through its field force — but almost no digital footprint aimed at the broader awareness layer around the condition its product supported, and its marketing team had never seriously considered any consumer-facing paid channel given the strict boundary against direct-to-consumer prescription promotion.
The business problem the brand brought to ICG wasn't about the product directly — it was about a gap earlier in the patient journey. Oncology patients and their families, in the period between a cancer diagnosis and the start of treatment-related supportive care, were searching extensively for information about what to expect, what supportive-care options existed, and which specialists managed that side of treatment — and almost none of that search behaviour connected back to any resource the brand's medical-affairs team had built, because the brand had never had a compliant channel for reaching that audience before treatment decisions were made. The oncologists prescribing the product, meanwhile, were themselves increasingly using conversational AI tools to cross-check supportive-care protocol questions and recent trial data between patient consultations.
The brand's medical and regulatory affairs team set the working boundary before any campaign planning began, and it shaped every subsequent decision: nothing in the patient-facing layer of the account could name the product, describe its mechanism, reference its dosing, or in any way function as promotion of a prescription medicine to a patient audience — the patient-facing layer could only ever educate on the condition and the supportive-care landscape broadly, pointing patients back toward a conversation with their own treating oncologist. A separate, clearly distinct HCP-facing layer, following full UCPMP 2024 discipline, could reference the product's clinical data directly, but only within conversations the account could reasonably determine were coming from a healthcare professional rather than a patient or caregiver.
The ChatGPT Ads campaign structure ICG designed
ICG built the Scale-tier account as two segmented conversation buckets within a single structure, each with its own bidding logic, creative set and compliance review standard, connected only by shared reporting at the account level. The patient-and-caregiver-facing bucket bid into disease-awareness conversations — questions about the specific treatment-related complication the product's category addresses, what supportive care generally involves, and how patients typically discuss these options with their oncology team — with every ad variant staying at the level of general medical information a patient advocacy organisation might publish, never referencing the brand's product by name or implying any specific therapy as the featured option.
The HCP-facing bucket bid into a narrower set of conversations that carried clearer professional signals — clinical-terminology density, protocol-specific phrasing, references to trial data or dosing considerations that a patient or caregiver would be unlikely to raise unprompted — and this bucket's creative could reference the product directly, including clinical-evidence summaries and links to the brand's medical-affairs portal, following the same UCPMP-clean review discipline ICG applies to any HCP-directed Rx creative. Conversation-stage bidding within the HCP bucket favoured mid-to-late-intent professional queries specifically, since early-exploration clinical questions from oncologists were judged less likely to represent a meaningful engagement opportunity than a more specific protocol or evidence question.
Landing infrastructure mirrored the two-bucket split precisely: the patient-facing layer routed to a condition-information microsite with no brand mention beyond a small, compliant corporate-sponsorship disclosure at the page footer, and a closing pointer toward finding and speaking with an oncologist rather than any product-specific next step. The HCP-facing layer routed to a professionally-gated medical-affairs portal, with a lightweight designation check ahead of clinical-evidence access, mirroring the verification approach ICG has used on other HCP-directed accounts, and offering a direct medical-information-request contact route rather than any transactional next step, since the product itself is prescribed, not purchased online.
A deliberate design choice worth naming: ICG did not attempt to route patient-facing conversation volume toward the HCP bucket's product information under any circumstance, even when a patient conversation showed strong downstream interest — the two buckets stayed structurally separate throughout the account's life, with the only bridge between them being the patient-facing content's generic encouragement to raise supportive-care questions with a treating oncologist, who might independently already know of or be reached separately by the brand's own field force.
The compliance discipline
DCGI/UCPMP 2024 governed the HCP-facing bucket in full — every clinical claim traced to a citable, peer-reviewed source, no comparative superiority claims against named or implied rival supportive-care therapies, and no promotional pattern that could read as incentivised engagement. The brand's own medical-affairs and regulatory teams reviewed every HCP-facing ad variant alongside ICG's compliance pass, a double-review structure the brand insisted on given the product's prescription-only status and the reputational sensitivity of any oncology-adjacent marketing misstep.
The patient-facing bucket's governing discipline was, if anything, stricter in a different direction: the working rule ICG applied was that no ad variant, landing page sentence or conversational excerpt could be read, even generously, as an attempt to influence a treatment decision toward a specific product. This meant rejecting draft copy more than once during setup that described the supportive-care category in terms specific enough that a reader familiar with the space could reasonably infer which product class was being referenced — a level of caution the brand's regulatory lead described as appropriately conservative for a first digital foray into this territory, even though it meant some awareness-stage copy read as more generic than the marketing team's initial draft had wanted.
ASCI Chapter III applied across both buckets as a secondary layer, and DPDP 2023 governed both funnels' data handling distinctly — the patient-facing microsite collected no personal data beyond anonymous analytics, deliberately avoiding any form that could create a sensitive-health-data collection obligation given the emotionally difficult context of oncology-adjacent search behaviour, while the HCP-facing portal's designation-verification step ran under the same minimal-collection, explicit-consent standard ICG applies to its other professional-gated funnels. The brand's data protection officer signed off on both flows before launch, a step ICG built extra review time for given the product category's sensitivity.
The 90-day outcome pattern
In the pattern ICG has observed across comparable Scale-tier dual-bucket Rx oncology accounts, the patient-facing awareness bucket ran at meaningfully higher conversation volume than the HCP bucket throughout the 90-day window, consistent with the much larger population of patients and caregivers searching this territory relative to the smaller pool of treating oncologists — illustrative figures put patient-facing completed conversations at roughly 90-120 a week by the second month, against an HCP-facing bucket running at 10-15 a week across the same period.
The patient-facing bucket's tracked outcome — a click-through to the condition-information resource or an external specialist-locator link — ran at a completion rate the brand's marketing team found encouraging relative to their prior (limited) experience with generic display advertising in adjacent categories, though ICG was careful throughout the engagement to frame this bucket's value as long-cycle brand and category awareness rather than anything resembling a direct sales metric, given the total absence of any product mention in this layer. The HCP-facing bucket's tracked outcome — a medical-information-request submission — ran at low absolute volume, as expected for a narrow specialist audience, but illustrative figures put 60-70% of these requests as ones the brand's medical-affairs team classified as genuinely new contacts, meaning oncologists the field force's existing coverage map hadn't already reached.
In GA4, both buckets aggregated under the AI Assistant channel, and the brand's analytics team ran a rough attribution check against its broader condition-awareness content elsewhere on its corporate site, finding AI Assistant referral traffic converting to the specialist-locator click-through at a materially higher rate than the brand's existing organic search traffic to the same resource — consistent with the pattern ICG has observed elsewhere of AI-assistant-referred traffic arriving with a more specific, further-along question already in mind than typical organic search traffic to the same page.
By the 90-day mark, the brand's medical-affairs team reported that three of the new HCP contacts generated through the account had led to a scheduled medical-information call, and the brand's regulatory review, conducted at the same checkpoint, found no ad variant across either bucket had required a takedown or revision — a result the regulatory lead specifically credited to the double-review structure and the conservative stance taken on the patient-facing bucket's copy throughout setup.
What we'd do differently next time
Running the two buckets within a single account structure, connected by shared account-level reporting, worked adequately but created some avoidable friction — the patient-facing bucket's compliance review cadence and the HCP-facing bucket's needed a genuinely different rhythm (the patient-facing layer needed lighter, faster review once its conservative ground rules were established; the HCP-facing layer needed a slower, more deliberate clinical-accuracy pass on every variant), and running both under one account's reporting cycle meant the faster-moving bucket sometimes waited on the slower one. Separate account structures from day one would let each bucket run its own appropriate review cadence.
The patient-facing bucket's copy, reviewed conservatively enough to avoid any inference-of-product-reference risk, ended up reading somewhat more generic than the marketing team had originally hoped — useful, accurate, but not distinctly memorable relative to other condition-information resources a patient might find. A more deliberate content-design pass, focused on making the disease-awareness content genuinely more useful and specific (without naming the product) rather than defaulting to safely generic phrasing whenever a borderline claim was flagged, is worth investing in earlier next time.
Finally, the HCP-facing bucket's designation-verification step, while appropriately cautious, may have been slightly too conservative for a specialist audience this narrow — several legitimate oncologist enquiries dropped off at the verification step according to portal analytics, and a lighter-touch verification (relying more on the professional-terminology signal already present in the conversation itself) might have captured a few more of the borderline-legitimate professional enquiries without meaningfully loosening the account's compliance posture.
How this maps to your own vertical
If your Rx or Rx-adjacent brand faces the same structural constraint — a product that can't be marketed directly to patients, but sits within a condition area patients are actively researching — the dual-bucket pattern here is the template worth adapting: a genuinely product-free, disease-awareness patient layer and a fully separate, UCPMP-clean HCP layer, kept structurally apart rather than blended into one funnel that tries to serve both audiences at once.
The discipline that carries across any Rx category is the same one this brand's regulatory team insisted on from the first meeting: when in doubt about whether a piece of patient-facing copy could be read as promoting a specific prescription product, the safer read wins, even at the cost of the copy sounding more generic than a marketing team would prefer. That single rule, applied consistently, is what let this account run a full 90 days with zero compliance takedowns.
