ICG ChatGPT Ads Training Series — Week 4 of 5

Week 4 · Landing-page discipline + attribution wiring

Backed by App\Support\NamedExperts::get(). --}}

By Rohit Gupta · Business & Growth Lead, Ichelon Consulting Group

LinkedIn

What you'll learn this week

By the end of Week 3 you had a working conversational ad — a sponsored response that shows up inside a ChatGPT conversation when a hospital administrator, a clinic owner, or a diagnostics chain's marketing head asks a buying-intent question. That ad earns a click. What happens in the next four seconds decides whether that click becomes a lead you can act on, or a bounce you can't even measure. Week 4 is about that four-second window and everything downstream of it.

We are going to cover three things in depth: how ChatGPT's own crawler and answer engine read your landing page through Speakable schema, why Person schema on your landing page changes how AI systems judge your credibility before a prospect even fills a form, and how to wire GA4 and your CRM so that a lead generated from a ChatGPT conversation is tagged as exactly that — not folded into "organic" or, worse, "direct" the way most agencies' setups do it today.

If you run India healthcare marketing and you've ever asked "where did this lead actually come from," this week gives you the plumbing to answer that question with certainty rather than a guess based on which channel had the fewest gaps in its data.

Core concept 1 — Speakable schema and why your landing page needs to be answerable, not just readable

Most healthcare marketing teams in India still build landing pages the way they did for Google Search in 2019: a hero banner, three bullet points, a form, a footer. That page works fine for a human scanning with their eyes. It works badly for an AI system that arrives not to display the page but to extract an answer from it and relay that answer back to a user inside a conversation.

Speakable schema (schema.org's SpeakableSpecification) was originally built for voice assistants, but it has become one of the clearest signals conversational AI systems use to identify which sections of a page are safe to quote verbatim inside an answer. When ChatGPT's crawler indexes your landing page and later decides whether to reference it in a follow-up answer to the user who just clicked your ad, it is looking for tightly-scoped, factually self-contained passages — not marketing copy that requires the rest of the page for context.

In practice, this means restructuring your above-the-fold content into short, declarative units. Instead of "We're India's trusted partner for hospital growth," a speakable-friendly equivalent reads: "ICG has managed patient-acquisition campaigns for hospital chains across 12 Indian cities since 2019." The second version is a fact. The first is a claim wrapped in adjectives. AI systems extract facts far more reliably than they extract sentiment, and every fact you make extractable is a fact that can travel back into a ChatGPT conversation as a citation — which is a form of reach you get with zero additional media spend.

The mechanical part is straightforward: mark your CSS selectors for the headline, the credibility line, and the primary value proposition inside a SpeakableSpecification block, referencing them by CSS selector path in your JSON-LD. The harder part is the editorial discipline — writing those specific blocks as clean, source-able facts rather than brand voice. We'll walk through a working example in the worked-example section below, but the principle to carry into your own landing pages is this: write the top of the page as though a machine will read it aloud to someone who cannot see it. If it survives that test, it will perform better in every AI-mediated channel, not just ChatGPT Ads.

There's a second, quieter benefit. Pages with well-structured Speakable markup tend to earn more consistent citations in AI Overview-style answers even outside paid placements, because the same structural discipline that makes a page speakable also makes it easier for any extraction-based system — Google's AI Overviews, Perplexity, Bing Copilot — to lift a clean answer from it. You're not building a ChatGPT-specific asset. You're building an AI-legible asset that happens to matter most, right now, for ChatGPT Ads because that's where the auction dollars are going.

Core concept 2 — Person schema: why the human behind the page matters more in AI answers than in search results

Google Search has spent a decade training marketers to think about E-E-A-T at the level of the domain and the article. ChatGPT's answer engine, by contrast, appears to weight the credibility of a specific named person disproportionately when deciding how much confidence to assign a claim before surfacing it in conversation. This is the practical reason Person schema on your landing page is no longer optional if you're running ChatGPT Ads — it's a direct input into whether the AI trusts what your page says enough to relay it.

Concretely: every landing page tied to a ChatGPT Ads campaign should carry a Person schema object for the named expert or founder associated with that service line, connected via author or reviewedBy to the page's primary content, and cross-linked to that person's LinkedIn profile, their bio page on your site, and — where applicable — their credentials. This isn't a cosmetic addition. It gives the AI system a verifiable entity graph to check against, rather than an anonymous brand voice it has no way to corroborate.

We see this pattern work best when the same named person appears consistently across a service line — the same author bio, the same headshot, the same credential set, repeated across every page in that cluster rather than rotated. Consistency here isn't a branding nicety; it's what lets an AI system build confidence in an entity over multiple encounters with your content, the same way a human builds trust in a source by seeing it get things right repeatedly rather than once.

A second, less obvious point: Person schema should be paired with genuinely first-hand claims. If your landing page says a named consultant "has advised over 40 hospital groups," that claim needs to be true and, ideally, traceable to a case study or testimonial elsewhere on your domain that an AI crawler can cross-reference. AI systems doing intent-matching for sponsored placements appear to penalize pages where the on-page claim and the supporting evidence don't line up — not through an explicit penalty mechanism you'll ever see documented, but through a lower confidence score that translates into your ad's response being paraphrased more cautiously, hedged more, or simply not surfaced as prominently in follow-up turns of the conversation.

The takeaway for Week 4: your landing page's credibility signal is no longer just your logo and your years-in-business line. It's a named, schema-marked, cross-referenced human being whose claims can be checked. Build that person's presence deliberately, and reuse them across every page tied to the campaign you're running.

Core concept 3 — GA4 and CRM wiring: making the ChatGPT Ads channel visible instead of invisible

Here is the uncomfortable truth most India healthcare marketing teams discover only after they've spent a few lakhs on ChatGPT Ads: without deliberate tracking setup, a meaningful share of the leads that channel generates get misattributed to "Direct" or "Organic Search" in GA4, because the referral pattern from an AI conversation doesn't always carry the clean UTM structure a paid search click does. If your reporting can't see the channel, your budget conversations with leadership will always undersell it — and an undersold channel gets cut, regardless of how well it's actually converting.

The fix has three layers. First, enforce UTM parameters on every ChatGPT Ads placement — utm_source=chatgpt, utm_medium=cpc, utm_campaign set to your specific service line — and verify at the platform level that these parameters survive the click-through rather than getting stripped. Second, build a custom channel group in GA4 that explicitly captures chatgpt as a source rather than letting it fall through to the default "Unassigned" or "Direct" bucket; this is a five-minute configuration change in GA4 admin that most agencies simply never make. Third, and most important, tag every lead event — form submit, WhatsApp click, callback request — with the session's attribution data before it ever reaches your CRM, because CRMs downstream of a poorly configured GA4 inherit its blind spots permanently.

Once that plumbing is in place, the number that matters starts to show up cleanly: across ICG's own tracked engagements, leads attributed to the "AI Assistant" channel in GA4 convert to a qualifying key event at roughly double the rate of organic search leads and around seventeen times the rate of leads bucketed as direct. That gap is not a fluke of small sample size — it reflects a structural difference in intent. Someone who asked a specific, comparative question inside a ChatGPT conversation and then clicked through to your page has already done more qualifying work on themselves than someone who typed your brand name into a search bar out of habit.

On the CRM side, the wiring you want is a webhook or API integration that passes the full attribution string — source, medium, campaign, and, where available, the actual query text that triggered the ad — into a custom field on the lead record at the moment of creation. This single field becomes the fastest way for your sales or BD team to open a call with context: "I saw you were comparing patient-acquisition options for a multi-city diagnostics chain" lands very differently than a cold "how can I help you today." Build this once, correctly, and every subsequent lead from the channel arrives self-documented.

Worked example — an anonymised ICG engagement pattern

Exercise for you to do this week

Common failure modes to avoid