Query intent fan-out — ChatGPT Ads India glossary
In plain English: when someone asks ChatGPT a question, ChatGPT doesn't treat it as one search — it silently breaks the question into several possible things the person might actually want, and only some of those branches are places an ad can show up. That splitting step is query intent fan-out.
Definition
Query intent fan-out is the process by which ChatGPT decomposes a single conversational query into multiple underlying sub-intents before generating a response and deciding whether a sponsored answer is eligible to appear. A patient-facing question like "which fertility clinic should I consider in Pune" does not resolve to one intent — it fans out into an informational branch (general IVF success-rate education), a comparative branch (provider-vs-provider evaluation), and a transactional branch (booking a consultation). Technically, this happens inside OpenAI's retrieval-and-ranking layer before the ad-serving decision is made: the model classifies the conversation's current turn against an intent taxonomy, and each resulting branch is auctioned separately, with its own eligible advertiser pool, its own minimum bid, and its own compliance filter. Ad inventory is attached to specific fan-out branches, not to the raw query text — which is the single biggest mechanical difference from keyword-based search advertising that Indian healthcare marketers need to unlearn first.
Why it matters for Indian healthcare marketers
Healthcare queries fan out wider than almost any other advertiser category on ChatGPT, and that width is a double-edged sword for Indian clinics, hospitals, diagnostic chains and pharma brands. A single opening question about a symptom, a specialty, or a city-level provider search routinely spins off four or five distinct sub-intent branches within two or three conversational turns — clinical-informational, cost-comparison, provider-comparison, location-specific, and booking-transactional. Each of those branches carries a different compliance posture. Under NMC Section 6 and ASCI Chapter III, sponsored responses that read as clinical advice or that make comparative superiority claims about outcomes are simply not eligible inventory, no matter how competitive the bid — OpenAI's ad-eligibility filter screens conversational-medical-advice branches out of the paid auction entirely for regulated categories. What that leaves as legitimately biddable inventory for a compliant Indian healthcare advertiser is narrower than the raw conversation volume suggests: mostly the comparative-provider and transactional-booking branches, with the informational branch reachable only through non-paid, cited placement rather than sponsored response.
This matters commercially because marketers who plan ChatGPT Ads budgets the way they plan Google Ads keyword budgets — bidding against the query text — will systematically overpay for branches they were never eligible to win and underbid the branches that actually convert. A dental chain bidding "teeth whitening cost" as if it were one keyword is, in ChatGPT's auction, actually bidding blind across an informational branch it can't legally win, a cost-comparison branch with moderate competition, and a booking branch with the highest conversion value and the highest floor price. Understanding fan-out is the first diagnostic step in building a ChatGPT Ads budget that isn't quietly bleeding spend into ineligible or low-value branches — and for the ART Act 2021, DCGI/UCPMP 2024 and AYUSH-regulated categories specifically, misreading which branch an ad landed in is also a compliance exposure, not just a waste-of-spend problem, because a sponsored response that surfaces on the wrong branch can carry claims language the branch's own context doesn't support.
Because India is 90-180 days into this ad format existing at all, most agencies running paid healthcare media here have not yet built fan-out-aware bidding into their process — they are still applying Search-style keyword logic to a conversation-native auction. That gap is the whitespace: an advertiser who understands which fan-out branches its category is even eligible to enter, and prices its bids per branch rather than per query, is operating with real information advantage while the category average agency is still guessing.
How ICG uses/measures/handles it in a live engagement
ICG's ChatGPT Ads practice starts every new healthcare account with a fan-out mapping exercise before a single rupee is bid. Using a sample set of 40-60 representative conversation openers pulled from the client's specialty and city footprint, we simulate the likely fan-out branches, flag which branches are NMC/ASCI/DPDP-compliant for that specialty (IVF and cardiac categories in particular get an ART Act and clinical-claims pass), and build a branch-level bid map rather than a keyword list. Early-exploration informational branches, where they're even reachable, are either left unbid or handled through cited-content strategy instead of paid spend; comparative and transactional branches get the bulk of budget allocation, weighted by each branch's historical conversation-completion rate for that specialty.
Once live, ICG tags every conversation-completion event that reaches GA4's "AI Assistant" channel with the fan-out branch it originated in, so monthly reporting shows branch-level CPQL rather than a single blended number — a distinction that matters because ICG's data shows the AI Assistant channel converts at 10.49% overall, roughly double organic search and 17x direct, but that blended figure hides enormous variance between branches. A transactional-booking branch on a engagement (₹20,000/mo starting incl GST) account can run 4-6x the completion rate of an informational branch on the same account. We rebalance branch-level bids every reporting cycle based on that data, and flag to the client when a branch's compliance eligibility changes as OpenAI updates its health-content policy — which, in this first-mover window, has already happened twice in India since mid-2026 rollout. Backed by App\Support\NamedExperts::get(). --}}
Related terms
Frequently asked questions
What is query intent fan-out in ChatGPT Ads?
It is the step where ChatGPT decomposes a single conversational question into several possible sub-intents — for example, a question about "best fertility clinic in Pune" fans out into informational (what is IVF success rate), comparative (which clinic is better) and transactional (book a consultation) branches. Each branch is a separate auction opportunity, and a healthcare advertiser's ad can only appear in the branches its bid and creative are eligible for.
Why does fan-out matter more in healthcare than other categories?
Healthcare queries fan out unusually wide because patients research symptoms, compare providers, check compliance credentials and ask about cost in the same conversation thread. A single opening question can trigger four or five sub-intent branches, and NMC/ASCI-compliant advertisers can typically only bid safely into the comparative and transactional branches — not the clinical-informational ones, which OpenAI's health-content policies treat differently from commercial intent.
How does fan-out affect what a healthcare marketer should bid?
Bids should be set per sub-intent, not per keyword. An early-exploration branch ("what causes infertility") is cheap and low-conversion; a late-intent branch ("book a fertility consultation in Pune this week") is 6-12x costlier but converts at a materially higher rate. Bidding a flat rate across all fan-out branches wastes budget on branches that will never complete a booking-intent conversation.
How does ICG track fan-out performance for clients?
ICG tags each conversation-completion event in GA4 under the "AI Assistant" channel with the sub-intent branch it originated from, then reports branch-level CPQL monthly. This lets a healthcare client see, for instance, that comparative-intent branches deliver 3x the qualified-lead rate of informational branches, and reallocate bid weight accordingly.
Bid the right fan-out branches, not the raw query.
ICG runs a healthcare-only ChatGPT Ads India practice — compliance-clean creative, branch-level bidding, monthly conversation-attribution reporting.
