How an ENT hospital ran Enterprise-tier ChatGPT Ads for its cochlear-implant programme across 3 states
The situation
Picture a multi-hospital ENT group operating flagship cochlear-implant programmes at three centres spread across three states, each centre staffed with its own implant surgeon, audiology team and post-implant rehabilitation unit. The group had built genuine clinical depth over a decade — high implant volumes, strong outcome data, established referral relationships with paediatric hospitals and audiology clinics — but its marketing had never matched that depth. Each centre ran its own disconnected local advertising, and the group had no unified way to reach the two very different populations who actually drive cochlear-implant volume: parents of infants and young children flagged through newborn hearing-screening or early developmental concern, and adults experiencing progressive or sudden hearing loss later in life.
These two candidate populations research almost nothing alike. A parent's search behaviour after a screening referral is urgent, anxious and compressed into days or weeks, often starting with a conversational query like "our baby failed the hearing test, what happens next" rather than a branded search for any specific hospital. An adult candidate's journey is typically slower and more comparison-driven, spanning months of research into implant versus hearing-aid tradeoffs, surgeon experience, and financing before ever contacting a centre. The group's flat, undifferentiated marketing treated both journeys identically, and leads arriving through that approach converted poorly because the messaging matched neither audience well.
Layered on top of the audience split was genuine state-level variation. Referral density from audiology networks differed by state, awareness of government hearing-health schemes differed by state, and even the typical financing conversation — insurance, EMI, scheme eligibility — played out differently in each market. A national campaign structure that ignored these differences would waste spend chasing conversation patterns that simply didn't exist in a given state. This was the case for Enterprise tier: three states, two structurally distinct candidate journeys, and a compliance profile complicated by paediatric-consent sensitivity that a smaller single-bucket campaign had no way to carry.
Campaign structure
Enterprise tier's architecture let ICG build this account on two intersecting axes rather than one. The first axis split every conversation bucket by candidate type — paediatric-candidate conversations and adult-candidate conversations were never blended, because their bid economics, landing content and compliance review requirements diverged too much to share a bucket. The second axis split each of those by state, giving the account six primary buckets in total, each carrying its own bid ladder calibrated to that state's specific mix of referral-driven versus self-initiated conversations.
Within the paediatric buckets, the bid ladder weighted heavily toward conversations carrying screening-referral or early-developmental-concern language — "failed hearing screening," "not responding to sounds," "audiologist recommended evaluation" — bidding these at the top of the ladder, 8-10x baseline, because parental urgency in this segment translated reliably into fast evaluation bookings once a centre's programme was presented clearly. General curiosity conversations about cochlear implants as a concept, not tied to a specific referral or diagnosis, sat at a much lower rung, since this traffic converts slowly if at all and Enterprise-tier budget discipline still required defending cost-per-qualified-evaluation even at scale.
Within the adult buckets, the ladder instead rewarded comparison-stage signals — conversations referencing hearing-aid versus implant tradeoffs, surgeon-experience questions, or financing and insurance-coverage questions — since these markers indicated a candidate far enough along in their research to book an evaluation within weeks rather than months. Early-stage adult curiosity conversations, still exploring whether hearing loss was even implant-eligible, sat lower on the ladder and were served lighter, informational ad variants rather than a full evaluation-booking push.
Landing infrastructure ran six dedicated pages, one per state-candidate-type combination, each stating that state's specific centre location, surgeon credentials, typical evaluation-to-surgery timeline, and a state-appropriate financing and scheme-eligibility summary reviewed against that state's actual scheme rules rather than a generic national statement. Paediatric pages led with what a first evaluation visit involves and what parents should bring, written to reduce first-visit anxiety; adult pages led with candidacy criteria and surgeon experience data. Every page carried machine-readable structured data describing the programme, procedure type and location, matched to how conversational assistants were observed pulling structured facts into their responses. Enquiry forms captured candidate type and referral source at the point of submission, feeding directly into the group's CRM so each of the three centres' front-desk teams received leads pre-routed to the correct intake workflow.
Reporting ran state-by-state and candidate-type-by-candidate-type in a monthly Enterprise-tier dashboard, reconciled against the group's CRM booking data centre by centre, so the hospital group's marketing leadership could see clearly which of the six buckets was performing and which needed adjustment rather than reading one blended national number that would have hidden real variation between states.
Compliance discipline
NMC Section 6 and ASCI Chapter III governed every ad variant across all six buckets, with particular weight on two specific risks this category carries more than most. First, outcome-guarantee language: cochlear-implant results vary by candidate, age at implantation and rehabilitation engagement, and no ad variant across the full account implied guaranteed hearing-restoration outcomes or specific speech-development timelines, even in paediatric messaging where parental hope makes this kind of overreach tempting to lean into. Every claim about outcomes was framed against the group's own published outcome data with appropriate qualification rather than as a promise.
Second, financing and scheme-coverage claims: several government and insurance schemes partially cover cochlear-implant procedures, but eligibility varies by state, income category and specific scheme rules, and ICG's compliance review rejected any ad language implying automatic or universal coverage, requiring instead that financing messaging point to "scheme eligibility varies, we'll confirm during evaluation" framing. This mattered doubly because a family under financial stress acting on an inflated coverage claim represented exactly the kind of harm ASCI's healthcare-advertising guidance is built to prevent.
Paediatric-candidate messaging carried a third layer of review specific to parental-consent framing — no ad copy created urgency pressure around a child's developmental window in language that could push a parent toward a rushed decision, and every paediatric landing page stated plainly that a formal audiology and ENT evaluation, not the ad itself, determines candidacy. DPDP 2023 governed all six enquiry forms, with consent language specific to handling a minor's health information where the paediatric forms were concerned, reviewed separately from the adult forms' standard consent language given the added sensitivity. Across the full 90-day run and all six buckets, every ad variant cleared compliance review before launch, with zero post-launch flags across any state.
90-day outcome pattern
In the pattern ICG has observed on comparable Enterprise-tier multi-state hospital accounts, month one ran as a genuine calibration phase across all six buckets simultaneously — Enterprise tier's budget allowed faster initial data accumulation than a smaller account would see, but six distinct buckets each still needed enough conversation volume before bid ladders reflected real behaviour rather than launch-week estimates. Illustrative combined volume across all three states in month one: roughly 45-60 conversation-completion events a week, split unevenly across buckets, with the paediatric buckets converging on stable bid ladders faster than the adult buckets given the sharper, more consistent urgency signal in referral-driven paediatric conversations.
By month two, combined weekly volume rose to an illustrative 90-110 conversation-completion events, and evaluation-booking conversion within 14 days — the relevant window given how compressed paediatric urgency tends to be — climbed from an illustrative 20-25% in month one to 38-42% in month two across the paediatric buckets specifically, while adult-bucket conversion, tracked on a longer 30-day window given the slower adult research cycle, moved from 12-15% to 22-26% over the same period.
By month three, combined weekly volume reached an illustrative 130-150 conversation-completion events across all six buckets, with state-level variation becoming clear enough to act on: one state's adult bucket materially outperformed the other two, traced back to that state's higher baseline audiology-referral density, prompting a mid-quarter budget reallocation toward that bucket's higher-performing bid ladder. Cost per qualified evaluation booking, blended across all six buckets, ran illustratively 30-40% below the group's prior disconnected per-centre advertising spend, with the paediatric buckets running notably more cost-efficient than the adult buckets given their sharper conversion signal.
In GA4, this traffic attributed to the AI Assistant channel across all three states, and the hospital group's own centre-level data showed a downstream pattern worth noting beyond the evaluation-booking numbers: a meaningfully higher share of paediatric-bucket evaluation bookings converted through to a scheduled implant surgery within the quarter compared to the group's historical conversion rate from evaluation to surgery booking, suggesting the campaign was reaching families further along in genuine decision-readiness rather than early-stage information-gatherers.
What we'd do differently
Running paediatric and adult conversations under one shared general-candidacy bucket for the first several weeks before the full six-bucket split matured cost the account some early efficiency — the paediatric urgency signal was strong enough that it should have been isolated into its own workstream from week one rather than allowed to blend with adult-candidate data during initial calibration, delaying how quickly the paediatric bid ladder reached its eventual high-performing shape.
The mid-quarter budget reallocation toward the strongest-performing state's adult bucket was the right call but came later than it should have — state-level performance divergence was visible in the data by week 5 or 6, and waiting until month three to reallocate meaningfully left efficiency on the table in the two weaker-performing states during that gap. A tighter review cadence, checking state-level divergence every two weeks rather than monthly, would have caught this sooner.
Financing and scheme-eligibility messaging, kept deliberately conservative for compliance reasons described above, was probably more conservative than strictly necessary on the adult-bucket pages specifically, where several state-specific schemes did have clearer, more statable eligibility criteria than the blanket "we'll confirm during evaluation" framing suggested — a more granular, state-specific financing statement on those pages, still compliance-reviewed, likely would have reduced hesitation at the enquiry stage without crossing any coverage-guarantee line.
How this maps to your own vertical
If you run a multi-location hospital programme spanning more than one state with structurally distinct candidate populations — paediatric versus adult, urgent-referral versus self-initiated, or any comparable split specific to your specialty — this two-axis bucket structure, splitting by both audience type and geography rather than defaulting to one national campaign, is the shape worth building toward regardless of whether your category is ENT, oncology, cardiology or another multi-state specialty programme.
The broader lesson is that Enterprise-tier value comes from the campaign architecture matching the real structural complexity of how your candidates actually decide, not from budget size alone — a hospital group running six well-differentiated buckets on Enterprise tier will consistently outperform the same total budget spread across one undifferentiated national bucket, and that structural discipline is worth investing in before scale, not after.
