I keep getting the same question from founders and CMOs lately.
“ChatGPT is running ads now — should we be doing something?”
The honest answer is yes, but almost certainly not the thing you’re picturing. Because the headline everyone’s reacting to — ads now appear in 51% of US ChatGPT replies — is true, and also a little misleading if you stop reading there.
Let me walk through what that number actually hides, and why India’s position in all this is stranger and more interesting than the agency checklists are letting on.
The number everyone’s quoting, and the five months behind it
Early July 2026: ads showed up in roughly 51% of US ChatGPT replies. Live in six countries. Big number, easy to screenshot.
But here’s the path that 51% took to get there.
- 🔹 April–May: a baseline of 0.42%. Basically nothing.
- 🔹 May 26: a spike to 26.5%.
- 🔹 June 14: a collapse to 0.05% — near zero.
- 🔹 Late June to early July: a rebound past 51%.
That’s not a channel that has found its footing. That’s a channel still being built while the plane is in the air.
And penetration isn’t even uniform across the six live markets. Canada sits at 53.6%, the US at 51.0%, Australia at 49.8% — but Japan is only 18.8%. There’s even a faint 3.6% trace showing up in the Netherlands, which isn’t officially on the list at all. (Test leakage or a quiet soft-launch? Nobody’s saying. I wouldn’t build a plan on it.)
So when someone tells you “ChatGPT Ads are a 51% channel,” the accurate version is: in one market, in one week, after a delivery graph that looked like a heart monitor.
What “six countries” means today
US, Canada, Australia, New Zealand, UK, Japan. Mexico, Brazil, and South Korea are described as “in progress.”
India? No official date. Anyone quoting you Q3 or Q4 2026 — and plenty are — is reading the expansion pattern, not an OpenAI roadmap. There is no announced India launch. There is no announced India price. Hold that thought, because it matters.
The pricing has already flipped twice — so treat any rate card as provisional
This is the part I wish more of the India-facing pieces would say out loud.
Look at how fast the money model moved at home:
- Feb 9, 2026 — pilot launches at a flat $60 CPM with a ~$200,000 minimum spend.
- Apr 10, 2026 — minimum cut to ~$50,000, CPC bidding introduced. Clearing CPMs reportedly fell to $25 in some categories.
- May 5, 2026 — full self-serve, no minimum spend, CPC bidding with $3–5 starting bids, Conversions API added.
- May–July 2026 — CPA bidding, then oCPC, geo-exclusions, and a bulk campaign API.
Two full pricing-model flips in three months. From a $200K club to a no-minimum self-serve auction.
Now here’s the problem. At least one widely-shared India guide still quotes the abandoned $60 CPM / $200K-minimum beta pricing as if it’s current, and pegs it at 6–30x typical display rates and 3–4x Meta’s pricing in India. That number was already dead by April.
I’ve written before about how frontier AI pricing moves — my Claude Fable 5 cost teardown made the same point. A new AI-native product prices fast and unpredictably before it settles.
👉 So if you’re modelling ChatGPT Ads for India, plan a range, not a point estimate. Whatever rate India gets quoted at launch will be a snapshot of a moving target, not gospel.
Why India’s “not live yet” window is structurally different
Here’s what almost nobody is connecting.
India is OpenAI’s second-largest market in the world. 100 million weekly active ChatGPT users — roughly 1 in 8 of ChatGPT’s ~800 million weekly users. Largest student user base of any country. And the base is overwhelmingly free tier.
Why does that matter? Because Free and Go are the only tiers ChatGPT Ads currently serves. Plus, Pro, Business, Enterprise, and Education users see no ads at all. (Go launched in India in August 2025 at ₹399 — about $4.50 — and OpenAI made it free for a year to Indian users that November. So the ad-eligible base here isn’t just free users, it’s free-plus-Go — and thanks to that free year, an even larger slice of India’s base is sitting in exactly the tiers where ads run.)
Sit with that for a second.
The pilot markets — US, Canada, Australia — are broadly wealthier, higher-subscription markets. That’s my read, not a number OpenAI publishes. But India’s base leans the other way, hard, toward exactly the Free-and-Go tiers where ads actually run. Which means the ad-eligible inventory here could dwarf the pilot markets once the switch flips. India might not be a “smaller market that gets ads later.” It could be the largest ad-inventory market OpenAI has.
And the local ad economy is already tracking this. WPP Media’s TYNY puts India’s total ad industry at ₹2,01,891 crore in 2026, up 9.7%, with digital at 68.1% share. Inside that, “intelligence-driven” formats — AI search, voice, agentic discovery — are forecast to grow around 8%. The category ChatGPT Ads lives in is already a line item in the forecasts.
So the window isn’t dead time. It’s setup time.
What’s actually worth doing in the window (and what isn’t)
Let me be practical, because “get ready” is useless advice without specifics.
Worth doing now:
- ✅ Build your entity and schema signals — consistent business data across every source an AI system reads.
- ✅ Test your organic visibility today in ChatGPT, Perplexity, and Google AI Overviews. Find where you’re invisible before there’s a paid auction to lose.
- ✅ Verify Bing Webmaster Tools — it feeds the retrieval layer behind these answers.
- ✅ Do budget scenario planning in rupees, in ranges, against those US pricing swings. Line up measurement now — Conversions-API-style tracking, and attribution that credits the organic mention that often precedes the paid click.
- ✅ Start tracking ad-adjacency as a metric. Is a competitor’s paid unit going to sit right next to your earned citation? Citation rate alone won’t tell you that.
Not worth doing yet:
- ❌ Building India-specific campaign tooling or API ops for a product with no India launch date, no India pricing, and a home-market delivery rate that swung from 51% to near-zero and back inside five weeks.
There’s a real trade-off here, and it’s easy to get wrong in both directions. Do nothing, and you enter India’s eventual auction cold, with thin entity signals — what one analysis nicely called “training data asymmetry,” where reputations built on word-of-mouth and physical presence are invisible to the models. Over-build, and you’ve poured ops budget into a channel that hasn’t stabilised even at home.
So what’s the window actually for?
It’s not a countdown to “spend the moment it opens.”
It’s a readiness test. The groundwork — entities, structured data, organic AI visibility, measurement — compounds quietly and puts you structurally ahead the day the auction opens. That part is worth starting today.
But the discipline that matters just as much is refusing to lock into today’s US rate card as if it’s the future. It’s already moved twice. The delivery mechanics are still unstable in the markets where it’s live. India will get its own numbers, on its own timeline, and you want to meet them with a plan built on ranges — not on a screenshot of a 51% that was never as solid as it looked.
Prepare like it’s coming. Budget like you don’t know the price. Because you don’t.
👉 Are you already building GEO and entity signals ahead of ChatGPT Ads reaching India — or waiting for the launch button? I’d genuinely like to hear how you’re thinking about the window. Drop your take below.