Silica Gel for Home
Target persona
Home organizer / consumer
Prompt context
Protecting valuables — books, clothes, leather, shoes — from humidity damage
Case study · LLM-native advertising
The First-Flight Advantage: ChatGPT Ads, Proven
How layered targeting — keywords, prompts and personas — turned a 9-day ChatGPT Ads pilot into a blueprint for AI-native performance media.
- Client
- A global leader in active packaging solutions
- Market
- India — Desiccant Solutions category
- Flight
- Aug 26 – Sep 3, 2026 · 9-day pilot
Executive summary · The pilot at a glance
111,698Impressions
4,090Clicks
3.66%CTR
$0.25Avg. CPC
$9.10CPM
$1,016.28Total spend
Headline read
The first ChatGPT Ads flight for this active-packaging brand drove 111,698 impressions and 4,090 clicks at a 3.66% CTR — well above typical benchmarks for social and display — for just $1,016.28 at a $0.25 average CPC and a $9.10 CPM. Efficiency held steady across the flight, validating ChatGPT as a scalable, cost-efficient performance channel for solution-specific demand.
01 — Pilot context
Why we tested a conversational AI surface.
Nine days live, a single ad account, four solution-specific ad groups — a deliberately small buy designed to answer one question.
Objective
Test whether conversational AI surfaces can drive qualified, in-market traffic ahead of a broader always-on allocation.
Channel
NeuGenM's LLM Ads Network — the brand's first LLM-native buy, sitting alongside Amazon Ads, Google Ads and Truecaller in the media mix.
Market & audience
India, CPC pricing, running against desiccant and moisture-control intent — silica gel, oxygen absorbers, container and pharma desiccants.
Flight
Aug 26 – Sep 3, 2026 (9 days live). Paused at close for this review before any next-flight decision.
02 — The NeuGenM difference
Layered context — not just keywords.
Google matches a query. Meta matches an interest graph. NeuGenM's LLM Ads Network layers three signals inside the live conversation itself.
01
Keywords
The category and product terms a user's conversation is actually built around — “silica gel”, “moisture control”, “container rain”.
02
Prompts
The specific question or task the conversation is solving — protecting a shipment, storing pharma stock, organizing a closet — not just a topic tag.
03
Personas
The end-consumer archetype behind that prompt — home organizer, logistics buyer, pharma packaging engineer — matched to the right solution ad.
The result: ads placed at the moment a real intent signal appears in-conversation — not against a static audience list.
03 — Campaign architecture
One ad group per consumer solution.
Structure followed persona, not just product SKU — each ad group was built around the specific problem an end consumer was solving.
Oxygen Absorber
Target persona
Packaged-foods brand owner
Prompt context
Preserving shelf life for sealed food packaging via iron-based O2 absorption
Container Desiccant
Target persona
Logistics / shipping buyer
Prompt context
Protecting cargo from “container rain” condensation across ocean transit
Pharma Desiccant
Target persona
Pharma packaging engineer
Prompt context
Safeguarding drugs, APIs and medical devices from hydrolytic degradation
04 — Delivery & efficiency
Delivery ramped, then held steady.
Volume built over the first two days, then settled into a consistent core flight before a partial final day.
Aug 26–27
Ramp-up
Initial delivery as the auction calibrated against the new placement.
Aug 28 – Sep 2
Sustained delivery
The core flight — roughly 13,000–18,000 impressions per day, the bulk of total volume.
Sep 3
Partial close
A shorter final day as the 9-day pilot window closed for review.
CTR held in a tight 3.5%–4.2% band across the flight
No erosion in click-through as delivery scaled — a sign the auction found a stable, relevant audience rather than exhausting it.
CPC compressed from $0.44 to $0.20–$0.25 as volume scaled
Consistent with normal auction dynamics during the Aug 28 – Sep 2 sustained window; CPC rose slightly on Sep 3's lower-volume partial day.
05 — Channel comparison
ChatGPT vs. Meta & Google.
This pilot's own results vs. 2026 published cross-industry benchmarks for Meta and Google Ads — not this brand's own Meta/Google spend.
Cost per click
Avg. CPC ($) · lower is better
Click-through rate
CTR (%) · higher is better
Read: at $0.25, this pilot's CPC ran roughly 65% below the Meta benchmark and 95% below the Google Search benchmark. CTR (3.66%) beat the Meta benchmark by nearly 2×; Google Search's CTR sits higher because it captures users already mid-search with explicit intent — a different funnel stage than an in-conversation placement.
Sources: LocaliQ 2026 Search Advertising Benchmarks; Meta Ads industry benchmark aggregators (Ryze AI, Digital Applied, Triple Whale), 2026.
06 — Audience insights
Where — and how — the pilot performed.
Delivery concentrated in the Mumbai and Delhi-NCR metros, on a near-even mobile/desktop split.
| City | Impressions | CTR |
|---|---|---|
| Navi Mumbai | 12,700 | 2.8% |
| Wadgaon | 6,856 | 2.8% |
| Gurugram | 3,867 | 2.6% |
| New Delhi | 3,585 | 1.9% |
| Mumbai | 2,685 | 2.7% |
| Bengaluru | 2,285 | 2.5% |
Device split
- 55%Mobile
- 45%Desktop
Near-even split — mobile reflects ChatGPT's own app usage; desktop reflects ChatGPT web usage in professional/research contexts.
Navi Mumbai and Wadgaon together delivered ~18% of impressions at a consistent ~2.8% CTR — the strongest pocket. A few city-level rows (e.g. an outlying “Mumbai City” CTR) warrant a data-quality check before reading them as a real signal.
07 — Key takeaways
What the pilot told us.
01
The media works
A 3.66% CTR and $0.25 average CPC are strong for a first-touch channel — ChatGPT surfaces are finding relevant, in-market attention for desiccant queries, at a fraction of Meta and Google's benchmark cost per click.
02
Layered targeting drove the efficiency
Structuring ad groups around keyword, prompt and persona layers — not just product SKUs — let placements surface at genuine moments of intent inside the conversation.
03
Delivery is geographically concentrated
Mumbai and Delhi-NCR metros drove the bulk of impressions at healthy CTRs; a few city-level anomalies warrant a data-quality pass before the next flight.
Case study · NeuGenM LLM Ads Network
The first-flight advantage is ready to scale.
Layered context — keywords, prompts and personas — let this pilot beat social-media CTR benchmarks at a fraction of Google and Meta's cost per click, on the very first flight.