ChatGPT Ads in India: What App Marketers Need to Know Before Spending ₹725

India has more than 100 million weekly ChatGPT users. Ads went live here on 29 August 2026, self-serve opened on 4 September, and the minimum daily campaign spend is ₹725.

Those three numbers have been quoted in every launch write-up published in the past week. Put together, they suggest a huge audience available at a trivial entry price. The maths underneath is less flattering, and if you run a mobile app, the gap between the headline reach and the reach you can actually buy is the first thing worth understanding.

This is a guide to what ₹725 a day buys in India, what you can and cannot measure on the channel, and how to structure a first test. Everything here is current as of September 2026, which on this platform means it has a shelf life measured in weeks.

Netting 100 million down to something you can buy

The 100 million figure is weekly active users, and weekly active users are not impressions.

Two adjustments matter. Roughly 85% of ChatGPT users sit on ad-eligible tiers, because Pro and Enterprise are ad-free. And fewer than 20% of users encounter an ad on any given day, because ads appear in a subset of conversations rather than on every response. Similarweb measured sponsored placements in 26% of ChatGPT responses in August 2026.

Stack those and the daily addressable pool in India is a fraction of the headline number, not a rounding-error fraction but a genuine order-of-magnitude difference. That is not a criticism of the channel. It is the same arithmetic that applies to every platform where the headline is monthly users and the purchase is daily impressions. It is worth doing before you present a forecast to anyone.

The number that actually constrains you is your budget against the prevailing CPM, and that is where India is genuinely unknown.

What ₹725 a day actually buys

No CPM data has been published for India. The only reference points are the US and UK, where CPMs fell from around $60 at the February launch to as low as $25 by April 2026 as inventory expanded.

India will probably clear lower, because auction pressure is lower and the market is new. How much lower is the open question. Here is the same ₹725 budget modelled across a range of CPMs, with a 0.50% click-through rate applied, which is the figure Similarweb measured on sponsored placements, and a 30% store conversion rate, which is a common mid-range assumption for app store listings.

These are illustrative figures built from published platform-level statistics, not measured India results. Substitute your own store conversion rate, which you already know, and the picture sharpens considerably.

The point of the table is not the specific numbers. It is that at floor budget, on any plausible CPM, you are looking at single-digit or low-double-digit installs per day. That has three implications.

It will take weeks, not days, to accumulate enough installs to say anything about quality. It means cost per install will swing wildly day to day on small samples, and you should resist reading those swings as signal. And it means a floor-budget test is a test of whether the channel works at all, not a test of how well it performs. Those are different questions and only the first one is answerable at ₹725.

If your total monthly spend is small enough that ₹725 a day is a meaningful proportion of it, the constraints in measuring ROAS when you are spending under ₹2L a month apply directly to how you should read these numbers.

Why the agency setup guides do not work for apps

Within days of the India launch, at least four Indian marketing agencies published ChatGPT Ads setup guides. They are competent and they are all about websites.

Every one of them describes the JavaScript pixel, and some cover the Conversions API. Both measure web conversions. Neither measures an app install, because an app install happens outside the browser, on a device the pixel never sees.

For a mobile app the route is different. Your campaign's destination has to be an attribution link rather than a store URL, the click identifier has to survive the redirect to the store, and your SDK matches the install on first open. OpenAI documents AppsFlyer and Adjust as its currently supported measurement partners, and attribution on this channel is click-based.

This distinction is the whole reason app marketers need separate guidance, and it is the most expensive thing to get wrong. A pixel-only setup on an app campaign produces a campaign that spends normally, reports clicks normally, and attributes nothing. The full mechanism is in how ChatGPT Ads attribution actually works, and the configuration steps are in the ChatGPT Ads install tracking setup guide.

Four measurement limits to plan around before you spend

Set expectations with your team before the campaign launches rather than after someone asks why a column is empty.

There is no cost API. Spend does not flow from ChatGPT Ads into your measurement platform automatically. ROAS for this channel has to be assembled by importing spend by hand. Budget the fifteen minutes a week, or accept that you are optimising on cost per install rather than return.

Reporting lags 24 to 48 hours. Attributed conversions do not appear in Ads Manager immediately. A day-one campaign will show spend against zero conversions, and that is normal rather than broken.

Targeting is more granular than reporting. You can target iOS, Android and web separately, but Insights returns two device buckets with mobile web folded into Mobile. India is an Android-first market, so this gap is more expensive here than almost anywhere else. Run separate campaigns per platform and split platform on your measurement side, where the SDK knows what the device actually was.

View-through is visible but inert. The one-day view-through window added on 19 August 2026 is excluded from bidding, billing and CPA. You can see the number. You cannot act on it.

None of these is disqualifying. All of them are reasons to treat the first month as a measurement exercise rather than a performance one.

What Indian app marketers should expect on quality

There is no India performance data yet, from OpenAI or anyone else. Nobody has published CPI, retention or ROAS figures for this market, and anyone who claims to has not shown their sample.

What can be said comes from adjacent markets and should be treated as a prior rather than a forecast.

Category composition skews commercial. Adthena found retail took 39% of ChatGPT ad placements while appearing on 24% of US queries, which suggests the platform surfaces ads disproportionately on transactional and comparison-shaped questions rather than across all conversation types. If your app sits in a category people research before choosing, that is favourable. If it sits in a category people discover socially, it is less so.

Advertiser density is still thin and heavily concentrated. Adthena counted 7,378 distinct advertisers in a single week, 60.1% of them in the US. India is entering an auction with little domestic competition, which is the main reason to expect low early CPMs and the main reason those CPMs will not stay low.

And the counter-evidence deserves stating plainly. Across 169,560 UK query scrapes in June 2026, Adthena recorded zero paid ChatGPT placements. A market being technically live does not guarantee inventory is being served in it at volume. Watch your own impression delivery in the first week before drawing conclusions from your cost per install, because a campaign that barely delivers will produce a cost per install figure that means nothing.

A first 30 days at floor budget

Structure the test so it answers one question at a time.

Days 1 to 3: prove the plumbing. Launch a single campaign with your attribution link in place. Do not evaluate performance. Confirm that clicks appear in your platform with populated click identifiers, that installs attribute to ChatGPT Ads rather than to organic, and that your chosen activation event fires against the channel. Wait the full 48 hours before concluding anything.

Days 4 to 14: establish a baseline. Hold budget flat at the floor and change nothing. You are collecting a sample, not optimising. Watch impression delivery specifically, because a channel that under-delivers is a different problem from a channel that delivers expensive installs.

Days 15 to 30: introduce one variable. Split by platform if you have not already, or test a second creative angle. One variable, not three. At this install volume you cannot read a multi-variable test, and the discipline in the five-day creative testing sprint applies with the timelines stretched to match the smaller sample.

Kill criteria, decided now rather than later. Write down before you launch what would make you stop. Reasonable candidates: impressions under a threshold you set for the first week, cost per install more than double your blended average after 30 days, or activation rate on ChatGPT Ads installs materially below your other channels. Deciding this in advance is the difference between a test and a slow budget leak.

How to fit this into an existing channel mix

The realistic case for ChatGPT Ads in India right now is incremental reach at low absolute cost, not replacement of anything.

At floor budget it will not move your blended numbers. What it can do is tell you, cheaply, whether a channel that is likely to matter in twelve months works for your app at all. That is a reasonable thing to buy for ₹725 a day, and an unreasonable thing to build a quarterly forecast on.

If you are formalising how a new channel earns budget against your existing mix, the method in the budget allocation framework for multi-channel app growth handles the general case, and the constraints in scaling campaigns without breaking unit economics become relevant only once the channel has proven itself.

One thing to watch as you add the channel: with ChatGPT Ads, Meta and Google all running, more than one network will occasionally claim the same install. Deduplication across networks matters more when you add a channel, not less, and a new channel with a small install count is exactly the one that gets over-credited or under-credited without anyone noticing.

Which Indian app categories have the best case for testing early

No India category data exists, so this is reasoning from platform behaviour rather than from measured results.

The pattern in the US data is that ads concentrate on transactional and comparison-shaped questions. Retail took 39% of placements while appearing on 24% of queries, which implies the platform surfaces sponsored results more readily when someone is evaluating options than when they are chatting.

Applied to Indian app categories, that favours anything people research before choosing. Fintech and insurance apps sit well here, because comparison questions about lending, investing and premiums are exactly the shape of query that attracts a sponsored answer. Marketplace and travel apps have a similar profile. EdTech is plausible for the same reason, though the purchase cycle is long enough that a one-day view-through window and a click-based model will understate the channel's contribution.

The categories with the weakest early case are those where discovery is social rather than deliberative. Gaming, short-form content and dating apps are typically found through feeds and word of mouth, and a research-shaped surface is a poor match. That does not mean the channel cannot work for them, only that it is not where the first ₹725 is best spent.

Whatever your category, the honest position in September 2026 is that this is a hypothesis to test rather than a rule to plan against. India has no published performance data for any vertical, which is exactly why your own 30 days of evidence is worth more here than in any established channel.

Frequently asked questions

How much do ChatGPT Ads cost in India?

The minimum daily campaign spend is ₹725. Effective CPMs have not been published for India by OpenAI or any third party as of September 2026. The only reference points are the US and UK, where CPMs fell from around $60 at the February launch to as low as $25 by April 2026.

Can I run ChatGPT Ads for an Android app in India?

Yes. Platform targeting supports iOS, Android and web separately. The constraint is on the reporting side: Insights returns two device buckets, with mobile web folded into Mobile, so Android performance is not separately visible in Ads Manager. Run separate campaigns per platform and split platform in your measurement stack instead.

Do the Indian agency setup guides work for a mobile app?

Not for install measurement. The guides published around the India launch describe the JavaScript pixel and, in some cases, the Conversions API. Both measure web conversions. An app install requires an attribution link from a measurement partner as the campaign destination.

Is ChatGPT Ads worth it for Indian apps right now?

At floor budget it is worth testing and not worth forecasting on. The channel is new enough in this market that no performance data exists, competition is thin, and the entry price is low. Those conditions favour a cheap test with written kill criteria, not a budget reallocation.

How do I compare ChatGPT Ads against Meta and Google?

On cost per install and activation rate first, since those are directly comparable. ROAS is harder, because spend does not flow automatically into your measurement platform and has to be imported by hand. If you need to justify the channel to a finance team, the method in proving attribution ROI to finance teams handles the general case.

Where Linkrunner fits

The test design above works on any measurement platform. What differs is how much of it you assemble by hand.

Linkrunner is the first MMP in India to go live with ChatGPT Ads, and attributes those installs alongside Meta, Google and TikTok in a single dashboard, without an enterprise contract or a minimum volume commitment. On a channel most Indian teams will test at ₹725 a day, a minimum commitment is the thing most likely to stop the test happening at all. Deduplication across networks runs in the same place, so the install three networks claim is credited once.

Setup is documented in the Linkrunner quickstart guide, and OpenAI's mobile measurement partner setup article is the primary source for how the channel behaves.

Key takeaways

• 100 million weekly Indian users nets down substantially: roughly 85% sit on ad-eligible tiers and fewer than 20% see an ad on a given day.
• At the ₹725 floor, expect single-digit to low-double-digit installs per day on any plausible CPM.
• Agency setup guides published for the India launch cover the web pixel and do not measure app installs.
• No cost API, a 24 to 48 hour reporting lag, two device buckets, and an inert view-through window are the four limits to plan around.
• Zero UK placements across 169,560 June scrapes is a reminder that a live market is not the same as a delivering one.
• Treat the first 30 days as a measurement exercise, with kill criteria written down before launch.

What to do this week

If you are going to test this channel, the highest-value work is not choosing an audience. It is making sure the installs it produces can be told apart from your organic traffic. That takes about twenty minutes and it is not recoverable retrospectively, because click-based attribution cannot be backfilled.

If you are not going to test it yet, the reasonable position is to wait for India performance data to exist. It does not yet, from anyone, and a channel with no published benchmarks is one where your own test is the only evidence available to you.

Either way, decide this week which of those two you are doing, rather than defaulting into a campaign that spends without measuring.

To see how ChatGPT Ads installs attribute alongside the channels you already run, request a demo from Linkrunner.