A ChatGPT Ads agency for brands buying the channel before the auction fills up

Yes. Ads began appearing to eligible ChatGPT users in India on 27 August 2026, initially through OpenAI's managed route and its named launch agency partners, WPP and Omnicom. Self-serve access to ChatGPT Ads Manager for Indian advertisers opens from 4 September 2026, with daily budgets starting at Rs 725. On 31 August 2026 OpenAI said advertisers across India, Europe, the Middle East and North Africa can buy ChatGPT ads directly through Ads Manager.

There are two costs. Media spend starts at a Rs 725 daily budget for Indian advertisers from 4 September 2026, which is roughly Rs 22,000 a month at the floor, though a floor budget will not produce a readable test. Management is scoped separately and GST applies on the management fee. Reported market benchmarks in mid-2026 put CPMs near $25 and recommended CPC starting bids around $3 to $5, but those are industry-reported figures rather than rates published by OpenAI.

Three differences matter. Targeting is by context hints, which describe the conversations and situations where your product fits, rather than by keywords and match types. The audience is limited to logged in adults on the ChatGPT Free and Go tiers. And the ad is a single chat card below the generated answer with a 50-character title and 100 characters of body copy, against the fifteen headlines and four descriptions a Google responsive search ad can carry.

Logged in adults on the Free and Go plans. Plus, Pro, Business, Enterprise and Edu subscribers have an ad-free experience, and ads are not shown to users identified or estimated to be under 18. No ads are served next to sensitive or regulated conversations including personal health, mental health and politics, and that exclusion applies regardless of the viewer's subscription tier.

No. OpenAI states that advertisers cannot pay to influence the answer ChatGPT generates. Ads appear below the response, carry a sponsored label and are visually separated from it. Changing what the assistant says about you is a separate discipline: answer engine optimisation and generative engine optimisation, which work on the sources the model retrieves and cites rather than on paid placement.

Context hints are short descriptions of the conversations, needs, topics or situations where your product or service is relevant, and they are set at the ad group level. OpenAI is explicit that they are not exact match keywords and that they do not guarantee delivery in any specific conversation. The practical consequence is that hints written as buyer situations tend to outperform hints written as product nouns, and there is no negative-keyword equivalent to fall back on.

For Indian advertisers, daily budgets start at Rs 725 from 4 September 2026. In the United States the minimum spend requirement that applied to the original pilot was removed when self-serve Ads Manager opened to all US advertisers on 5 May 2026. A platform minimum is not a sensible test budget, though: you need enough daily volume to clear a learning period before the numbers mean anything.

Political advertising is not permitted. Restricted and excluded categories include adult sexual content, gambling, tobacco and recreational drugs, weapons, illegal or dangerous goods, deceptive schemes and academic cheating services. Health and financial services advertisers may be eligible under stricter eligibility criteria and enhanced verification, and licensed businesses may be asked to evidence their credentials. Separately, no ads run next to sensitive user conversations even for approved advertisers.

Through the measurement pixel and Conversions API for web conversions, which went live with the US self-serve launch on 5 May 2026, plus UTM and source tagging that carries into your CRM so a lead is still identifiable at close. Platform reporting itself is aggregated: advertisers receive campaign performance data such as impressions and clicks, not user-level data, chats, memories or identities.

Yes, because they are separate systems. Paid placement puts a labelled card below the answer. Organic AI visibility determines whether your brand is named inside the answer itself, and that cannot be bought. A brand that only buys ads is absent from the recommendation and present only in the advertisement beneath it. The two channels are usually run together, with content and citation work feeding the organic side.

They can be, with two caveats. Assistant conversations skew heavily toward research and shortlisting, which is genuinely useful for considered B2B purchases, and context hints let you describe a buying situation rather than guess at a keyword. The caveats are that the audience sits on the Free and Go tiers, and that with no true CPA bid type yet you carry more of the measurement burden yourself. Treat it as a pipeline test with CRM-level tracking, not a lead-volume channel.

Plan for six to eight weeks before drawing conclusions. A relevance-weighted auction needs time to establish how your ads and landing pages score against real conversations, context hint iteration takes several cycles to settle, and the conversion-optimised objective needs conversion volume before it can bid usefully. Judging the channel at two weeks measures your setup speed, not the channel.