On September 4, OpenAI opens self-serve access to ChatGPT Ads Manager in India. Any business can then run campaigns directly, with daily budgets starting at ₹725.
This follows a managed rollout that went live on August 27. OpenAI launched with roughly 50 brands and agency partners WPP and Omnicom, showing ads to users aged 18 and above on the Free and Go plans. Subscribers on Plus, Pro, Business, Enterprise and Education tiers continue to see an ad-free product.
India was never going to be an exception. The ad product is already running in 38 countries, including the United States, Japan, South Korea and 31 European markets, and India represents more than 100 million weekly ChatGPT users.
Most of the coverage this week has been about CPMs, creative specs and who bought the first slots. That is the small story.
The line nobody is reading closely
Buried in OpenAI’s own communication is the detail that should reshape your 2027 budget.
Ads carry clear labels, appear beneath responses rather than alongside them, and stay visually distinct from the content. Advertisers cannot influence what ChatGPT actually says. The answers stay optimised for what is most useful to the person asking.
Read that again, because it defines two entirely separate pieces of real estate:
The answer. The recommendation itself. The three tools ChatGPT names when someone asks for the best project management software for a 40-person agency. This cannot be bought.
The slot below it. A labelled sponsored unit. This can be bought, starting September 4, by anyone with ₹725 a day.
Every brand that gets named in the answer is being handed distribution for free. Every brand that isn’t will be renting attention underneath a competitor’s recommendation, forever.
Why this is not Google all over again
The instinct is to treat this like paid search and wait. That instinct is wrong for one structural reason.
ChatGPT ads do not run on keywords. Advertisers define intent categories, and the system matches those against the full context of a conversation rather than against individual search terms. There is no long tail to hide in and no cheap corner of the auction to occupy while the big spenders fight over head terms.
Worse, the organic side is not a keyword game either. When someone asks a model to compare vendors, the model assembles an answer from what it has crawled, what it has been trained on, and what it can retrieve and cite in the moment. Ranking fourth on Google is not the same as being named in a three-item list generated in a single conversational turn. There is no page two in an AI answer.
So you have a paid channel with no cheap entry point and an organic channel that most brands are not measuring at all.
The cost of waiting, honestly
We should be precise here, because the scare-number version of this argument is dishonest.
₹725 a day is not expensive. That is the floor, not the ceiling. The real cost shows up in three places over the next eighteen months:
- Auction density. Right now roughly 50 advertisers are live in India. When self-serve opens on the 4th, that number goes up sharply, and every new entrant in your category bids against you for the same intent clusters. Prices go one direction.
- Permanence. Paid placement stops the day you stop paying. It leaves nothing behind. A well-structured, citable, crawlable content asset keeps getting pulled into answers long after it is published.
- Compounding. Brands that start building AI visibility now get named in answers for free while their competitors pay for the slot below. That gap widens with every month of head start.
At any real scale in a competitive category, a year of sustained ChatGPT ad spend runs into lakhs. That is not a reason to avoid the channel. It is a reason to make sure paid is doing incremental work rather than compensating for organic invisibility you never diagnosed.
What organic AI visibility actually requires
This is where most teams discover their SEO playbook does not transfer cleanly.
Are AI crawlers even reaching you? GPTBot, ClaudeBot, PerplexityBot and OAI-SearchBot hit your server every day, or they don’t. Most brands have never looked. A JavaScript-rendered site that Googlebot handles fine can be effectively invisible to an AI crawler that doesn’t execute your bundle.
Which pages are they actually pulling? Crawl volume is not the metric. The metric is which specific URLs get requested, how often, and whether those are the pages that carry your commercial argument.
Are you named, and where? The only question that matters commercially is whether your brand appears when a buyer asks a purchase-intent question, and whether you’re the first recommendation or a footnote in position seven. That requires tracking real prompts across ChatGPT, Gemini, Perplexity and Claude on a schedule, not spot-checking once a quarter.
Is your content structured to be quotable? Models extract claims. Pages built as narrative marketing copy give them nothing clean to lift. Pages built with direct answers, explicit comparisons, specifications and clear entity definitions get cited.
Can you tie any of it to traffic? AI referral clicks show up in your logs with identifiable referrers. If you are not separating that traffic, you cannot prove the channel works, and you will default to paid because paid has a dashboard.
The global lesson inside the India news
If you are reading this from London, Toronto or Sydney, your version of September 4 already happened. Ads went live for US users in February and expanded across Europe in August.
The sequence is identical in every market. Free tier gets ads. Agency-managed cohort goes first. Self-serve opens weeks later. Auction fills. Organic mentions become the only unpaid distribution left in the fastest-growing discovery channel of the decade.
India is simply the clearest live case study, because the self-serve gate opens on a date you can put in a calendar.
What to do in the next 30 days
Do not choose between paid and organic. Sequence them.
- Establish your baseline. Which AI crawlers reach your site, which pages they request, and whether your brand appears for the twenty prompts your buyers actually use.
- Fix crawl access before writing anything new. Rendering, robots directives, response codes.
- Restructure your top ten commercial pages to be extractable rather than merely persuasive.
- Separate AI referral traffic in analytics so the channel has a number attached to it.
- Then test ChatGPT ads with a real budget, against a baseline you can measure them against.
Teams that run this sequence buy paid placement as an accelerant. Teams that skip it buy paid placement as life support.
Frequently asked questions
Ads went live on August 27, 2026 through a managed rollout with roughly 50 brands and agency partners WPP and Omnicom. Self-serve access to ChatGPT Ads Manager opens on September 4, 2026, at which point any approved business can run campaigns directly.
Daily budgets start at ₹725. Advertisers choose a Reach objective and pay per thousand impressions, or a Clicks objective and pay per valid click, setting a maximum bid at the ad group level.
Users aged 18 and above on the Free and Go plans. Go costs ₹399 per month in India. Subscribers on Plus, Pro, Business, Enterprise and Education plans do not see ads. OpenAI also withholds ads from accounts identified or predicted to belong to under-18 users.
No. OpenAI states that advertisers cannot influence ChatGPT’s responses and that answers remain optimised for what is most helpful to the user. Ads are clearly labelled, appear beneath the response rather than inside it, and stay visually distinct from the content.
No. Advertisers receive aggregated campaign performance data such as impressions, clicks, spend, CTR and conversions. They do not get access to individual conversations, chat history, memory or personal information, and OpenAI states it does not sell user data to advertisers.
Advertisers define intent categories rather than keyword lists. The system matches those categories against the context of the live conversation, and where ads personalisation is enabled, select signals from the user’s broader ChatGPT experience. This is closer to audience and topic planning than to search term bidding.
Yes. Ads are not eligible near sensitive or regulated subjects, including health, mental health and politics.
ChatGPT ads are paid placements below an answer. AI visibility is whether your brand is named inside the answer itself. The first is rented and stops when the budget stops; the second is earned through crawlability, content structure and citation-worthiness, and keeps generating mentions without ongoing spend. Platforms like LLMAudit measure the earned side by tracking which AI crawlers reach your site and whether your brand appears when buyers ask purchase-intent questions.
Manual spot-checks are unreliable because responses vary by phrasing, session and personalisation. Consistent measurement requires running a fixed set of buyer-intent prompts on a schedule across ChatGPT, Gemini, Perplexity and Claude, then recording whether your brand is cited and in what position. LLMAudit tracks this as three states – not cited, cited in a low position, or cited among the top recommendations – so movement over time is visible rather than anecdotal.
The most common cause is client-side rendering. A React or Vue site that Googlebot processes without difficulty may return a near-empty HTML shell to an AI crawler that doesn’t execute JavaScript. Other causes are robots.txt rules blocking GPTBot or ClaudeBot specifically, aggressive bot protection, and content locked behind interactions. A server-side audit shows which crawlers reached which URLs and what status codes they received, which is the only reliable way to confirm access rather than assume it.
Yes. Static tracking parameters such as UTMs persist on ad clicks and can be read in your existing analytics. Earned AI traffic is harder – it arrives as referrals from chatgpt.com, perplexity.ai and claude.ai, and separately as crawler requests from GPTBot, ClaudeBot and PerplexityBot that never appear in GA4 at all. LLMAudit reads this at the server log level, which is where crawler activity actually shows up, and separates AI bot requests, AI referral clicks and human traffic into distinct streams.
Establish your organic baseline first, then test paid against it. Without a baseline you cannot tell whether ads are producing incremental growth or compensating for organic invisibility. Both channels are worth running; the sequence determines what you learn from the spend.
Is this only relevant to India?
No. The ad product is already live in 38 countries, including the United States, Japan, South Korea and 31 European markets. The rollout sequence has been consistent across markets, so the India timeline is a preview for anywhere the auction has not yet opened.
See where you stand
LLMAudit tracks server-side AI bot activity, monitors brand citations across ChatGPT, Gemini, Perplexity and Claude, and scores your pages on how easily AI systems can extract and cite them.
[Book a demo – https://www.llmaudit.ai/] and we’ll show you your current AI visibility baseline before the auction gets expensive.
