How a cardiac-imaging medical device company ran Scale-tier ChatGPT Ads targeting HCP-only intent
The situation
Picture an Indian medical device company with a cardiac-imaging portfolio — a portable echocardiography system and a cath-lab-adjacent intravascular imaging consumable line, sold into tertiary and secondary-care cardiology departments through a mix of direct enterprise sales and a regional distributor network. The company's existing demand generation ran almost entirely through medical conferences, distributor relationships and a modest LinkedIn presence aimed at cardiology department heads and hospital procurement leads. Digital advertising had never been part of the mix — the marketing lead's prior experience with broad-reach platforms had produced patient-facing enquiries the sales team had no use for, and the company had concluded, reasonably, that consumer ad platforms simply weren't built for a capital-equipment buyer.
The business problem was narrower and more specific than a typical healthcare ChatGPT Ads brief: the company wasn't trying to reach more people, it was trying to reach the right forty or fifty people a quarter — cardiologists evaluating a new imaging platform for their department, cath-lab administrators comparing consumable lines ahead of a renewal cycle, biomedical engineering leads researching technical specifications before a procurement committee meeting. Conference footfall and distributor introductions were producing enquiries, but slowly and unpredictably, and the sales cycle from first contact to a qualified demo request often stretched past four months with no visibility into what was happening in that gap.
What made ChatGPT Ads worth testing, in the scoping conversation, was the pattern the company's sales team had already noticed anecdotally: cardiologists and biomedical engineering leads were increasingly using conversational AI tools to research device specifications, compare imaging modalities, and draft procurement justification documents before ever picking up the phone to a vendor. That behaviour — professional, research-stage, happening well before a distributor call — was exactly the conversation-stage intent a Scale-tier HCP-only account was built to capture, provided the account excluded patient-facing symptom and diagnosis queries entirely, which the company's regulatory and clinical-affairs lead treated as a non-negotiable boundary from the first meeting.
The ChatGPT Ads campaign structure ICG designed
ICG built the account around a single Scale-tier conversation bucket scoped exclusively to clinical and procurement-framed cardiac-imaging queries, with an explicit negative-intent layer excluding any conversation that resolved toward a patient's personal symptoms, a specific diagnosis question, or "should I get an echo" phrasing — the kind of query a patient, not a cardiologist, would type. This exclusion layer was the single most important design decision in the account, because a cardiac-imaging device's ChatGPT Ads presence sits adjacent to a much larger volume of genuinely patient-facing cardiac-health conversation, and without a disciplined negative-intent filter, budget would leak toward an audience with no purchasing role at all.
Within the HCP-only bucket, ICG split conversation-stage bidding into three sub-bands reflecting the company's actual buying committee: a cardiologist-clinical band (conversations about imaging modality comparisons, resolution and diagnostic-accuracy questions, clinical-evidence requests), a procurement-technical band (conversations about consumable compatibility, service contracts, total-cost-of-ownership questions), and a biomedical-engineering band (conversations about technical specifications, installation requirements, integration with existing cath-lab infrastructure). Each band ran its own creative variant and its own bid ceiling, with the procurement-technical band carrying the highest per-conversation bid given its closer proximity to an actual purchase decision.
Landing infrastructure was built as a professional-gated microsite rather than an open consumer landing page — a short verification step (professional email domain check plus a designation dropdown: cardiologist, cath-lab administrator, biomedical engineer, procurement lead, other) sat ahead of the clinical-evidence library and the demo-request form, both to keep the funnel honestly HCP-only and to give the company's sales team a pre-qualified lead the moment a form was submitted. This verification step deliberately traded away some conversion-rate optimism for lead quality — ICG's experience across B2B medical device accounts is that an unverified, frictionless funnel produces a materially higher volume of unusable enquiries from students, patients and unrelated researchers than a lightly gated one does.
Bidding ran at a lower absolute conversation volume than a typical Scale-tier consumer healthcare account — this is expected and by design for a narrow B2B specialist audience — but at a meaningfully higher bid ceiling per conversation, reflecting the downstream deal value of a single qualified cardiology-department enquiry against a capital-equipment or multi-year consumable contract. Reporting ran monthly, reconciled directly against the company's CRM pipeline stages rather than against ICG's own funnel metrics alone, since the real measure of the account's value was how many verified HCP enquiries progressed to a scheduled demo, not how many conversations completed.
The compliance discipline
NMC Section 6 governed the boundary the account was built around from day one — Section 6 restricts how medical professionals and, by extension, entities marketing to or through them can be addressed in promotional content, and ICG's copy review treated every ad variant as if a practising cardiologist would read it critically, which is exactly who was meant to. This meant no claim of superiority over named or implied competing imaging systems, no unverifiable diagnostic-accuracy percentages without a citable clinical source, and no framing that suggested the device replaced clinical judgment rather than supporting it.
A specific discipline point that shaped the negative-intent exclusion layer: ICG's compliance review flagged, early in account setup, that even a well-intentioned HCP-framed ad could inadvertently surface in a patient-initiated conversation if the keyword and topic scoping wasn't tight enough — a patient researching "what is cardiac imaging" sits close, semantically, to a cardiologist researching "cardiac imaging modality comparison," and the platform's intent-matching needed explicit boundary-setting rather than being left to infer the difference from context alone. ICG tightened the topic scoping twice in the first three weeks after reviewing early conversation logs and finding a small number of patient-adjacent conversations slipping through the initial filter.
DPDP 2023 governed the professional-verification step's data handling — the email-domain check and designation dropdown collected the minimum information needed to gate access, with explicit consent language before any clinical-evidence download, and no retention of unverified attempts beyond a short window needed for fraud and abuse monitoring. Because the funnel never asked for or collected any patient health information, DPDP's most demanding provisions around sensitive personal data didn't come into play here in the way they would for a hospital's patient-enquiry form, but ICG still applied its standard healthcare-grade consent and retention discipline to the professional data that was collected, treating a cardiologist's or biomedical engineer's professional details with the same care it would treat any B2B healthcare buyer's data.
The 90-day outcome pattern
In the pattern ICG has observed across comparable Scale-tier HCP-only medical device accounts, the first three to four weeks ran at low, deliberately unhurried volume while the negative-intent exclusion layer was being tightened against real conversation logs — illustrative figures put completed HCP-verified conversations in the range of 8-12 a week during this calibration period, a volume the company's marketing lead initially found underwhelming until the exclusion-layer tightening work was explained as the reason the numbers were low but clean.
By weeks 5-8, with the three-band bid structure seasoned and the exclusion layer stable, verified HCP conversations rose to an illustrative 20-28 a week, with the procurement-technical band producing a disproportionate share of demo-request form fills relative to its conversation volume — consistent with the expectation that a procurement-stage conversation sits closer to an actual buying decision than an early clinical-curiosity one. Across the full 90-day window, illustrative figures put total verified HCP conversations at roughly 190-220, of which 22-28 resolved into a completed demo-request form, and of those, the company's sales team classified 14-17 as genuinely sales-qualified after their own follow-up call — a qualification rate meaningfully higher than the company's historical conference-lead qualification rate, which the sales lead attributed to the professional-verification gate filtering out the casually curious before a salesperson's time was spent on them.
Cost per verified HCP conversation ran substantially higher than a typical consumer healthcare Scale-tier account — expected, given the narrower audience and higher bid ceilings — but cost per sales-qualified lead compared favourably against the company's blended cost per qualified lead from conference attendance and distributor-sourced introductions over the same period, once conference travel, booth and staff-time costs were factored in by the company's own finance team. In GA4, this traffic aggregated under the AI Assistant channel; the company didn't have a large enough baseline of other paid-channel HCP traffic to run a like-for-like comparison, so this engagement's most useful benchmark was internal, against its own historical lead sources, rather than against another paid channel.
By the 90-day mark, two sales-qualified leads from the account had progressed to a scheduled on-site demo with a cardiology department, and the company's regional sales lead flagged both as departments that hadn't previously appeared on the distributor network's pipeline radar at all — a result the marketing lead described as the clearest early evidence that the account was surfacing genuinely new demand rather than simply capturing intent that would have reached the sales team through an existing channel anyway.
What we'd do differently next time
The professional-verification step was added to the landing flow roughly three weeks into the engagement, after the first batch of enquiries included a noticeable share of patients and students the sales team had no use for. Building that verification gate in from week one, rather than adding it after an early lead-quality problem surfaced, would have avoided burning early budget on unqualified enquiries and given the sales team a cleaner signal from the very first week.
The negative-intent exclusion layer needed two rounds of tightening in the first three weeks based on real conversation-log review, which is a normal part of any HCP-only account but did cost some early budget efficiency. A more conservative, narrower initial topic scope — deliberately erring toward excluding more borderline conversations at launch and widening carefully once the boundary was proven, rather than starting broader and narrowing down — would likely have produced a cleaner first month at a modest cost in early volume.
Finally, the three-band bid structure (clinical, procurement-technical, biomedical-engineering) was designed from the company's own understanding of its buying committee, which proved broadly right but slightly underweighted the biomedical-engineering band's actual influence on final purchase decisions — several sales-qualified leads showed engineering-band conversation history contributing meaningfully to committee sign-off even when the clinical or procurement band had opened the enquiry. A repeat engagement would weight the engineering band's bid ceiling closer to the procurement-technical band's rather than treating it as a lower-priority third tier.
How this maps to your own vertical
If your medical device or capital-equipment product is sold to a healthcare professional or institutional buyer rather than directly to a patient, the structural pattern here — an explicit negative-intent layer excluding patient-facing queries, professional verification ahead of your clinical-evidence or demo funnel, and bid bands mapped to your actual buying committee's distinct roles — is the template worth adapting, regardless of whether your device sits in cardiology, orthopaedics, diagnostics or another specialty entirely.
The discipline that doesn't change across device categories is the exclusion boundary itself: any B2B medical device account that doesn't deliberately filter out patient-facing intent will, over time, leak budget toward an audience with no purchasing authority and risk NMC Section 6 complications it never needed to court. Getting that boundary right in the first two to three weeks, before scaling spend, is the single highest-leverage decision in an HCP-only ChatGPT Ads build.
