TL;DR
- ChatGPT ads appear below a response, labelled as sponsored. OpenAI says they don't influence the answer and run on separate systems from the chat model.
- There are no exact-match keywords. You describe relevant conversations in "context hints", and an auction weighs relevance against your bid.
- Only Free and Go users see ads. Plus, Pro, Business, Enterprise and Edu accounts don't, and nor do accounts identified as under 18.
- Businesses in India can buy ads themselves through Ads Manager, but financial services and health ads are generally prohibited outside the US. Check this before you plan a budget.
- Billing is per impression or per valid click. OpenAI suggests starting click campaigns at $3 to $5 per click.
What a ChatGPT ad looks like
An ad sits under the end of a ChatGPT reply, separated from it and marked as sponsored. According to OpenAI's advertiser basics, each ad has six parts: your business name, your favicon, a title, a short description, an image and the landing page it links to. More than one ad, from different advertisers, can appear under the same answer.
That layout matters for how you write it. The person has just read an answer to their own question. Your ad competes with that answer for the next click, so it has to add something the answer didn't: a specific product, a price, a tool they can use now.
Who sees them
This shapes who you can reach. Per OpenAI's help page on ads in ChatGPT:
- Ads show to users on the Free and Go plans. Plus, Pro, Business, Enterprise and Edu accounts see none.
- Accounts identified as under 18 see no ads. OpenAI infers age from stated age, account age, usage patterns and topics, plus age prediction where available.
- Temporary Chats show no ads. Logged-out users only see ads suitable for all audiences.
- Free users can switch ads off by accepting lower message limits and fewer tools.
In practice, you're not reaching paying ChatGPT subscribers or most people using it at work on a company plan. For a B2B product whose buyers mostly use ChatGPT through a business account, that's a real gap in reach.
How an ad gets chosen
There's no keyword list. When a conversation is eligible for ads, the system considers four inputs:
- the context and intent of the current conversation
- your landing page
- your ad's title and copy
- the context hints you set on the ad group
If the user has ad personalisation on, it can also use signals from their wider ChatGPT use, including past chats and memories. With personalisation off, it uses only the current conversation and their general location and language.
Ads are then ranked in what OpenAI calls a relevance-weighted, second-price auction. A highly relevant ad can beat a higher bid, and the winner pays based on the next-best offer rather than its own maximum. Your landing page and copy are scored as part of that relevance, so a vague page costs you money, not just conversions.
Two things the system doesn't do. Advertisers can't shape, rank or change what ChatGPT says. And advertisers never receive conversations, chat history, memories, names, emails, precise location or IP addresses. You get aggregated numbers like views and clicks, plus any message a user sends you directly.
Context hints: the part most people get wrong
OpenAI is explicit that context hints are not exact-match keywords and don't guarantee delivery. They're descriptions of the conversations you want to appear in. Writing them like a Google Ads keyword list throws away the one advantage this channel has: it knows what the person is trying to do, not just what they typed.
Compare two sets of hints for the same product, an invoicing app for freelancers. This is an illustrative example, not a client campaign:
- Keyword thinking: invoicing software, invoice app, GST invoice, billing tool.
- Context thinking: freelancer asking how to send a GST-compliant invoice to a client; designer comparing ways to get paid faster by overseas clients; someone asking what to include on their first invoice.
The second set describes situations, which is how a conversation actually reads: people explain the problem they have, often without naming a product category. Write three to six hints per ad group, each one a real situation your product solves, and keep each ad group to one situation, so the copy and landing page can match it closely.
What you pay and how you bid
You choose one of three objectives:
- Views: you pay per thousand impressions, with a maximum CPM bid. Use it for awareness.
- Clicks: you pay per valid click, with a maximum CPC bid. Most first campaigns should start here.
- Conversions: the system optimises toward a conversion event you've set up. With click billing (oCPC), you still pay per valid click, not per conversion.
Bids are set per ad group. For click campaigns, OpenAI's own guidance is to start at $3 to $5 per click, and Ads Manager may show you whether a bid is likely to compete. That's a starting point from OpenAI, not a benchmark for your category, so expect to adjust once you see your own numbers.
What you can't advertise, and the catch for India
This is the section to read before you plan anything. OpenAI's ad policies prohibit these categories outright: gambling and betting, dating and adult content, alcohol and tobacco, cannabis and other recreational drugs, political ads, counterfeit goods, individual job and housing listings, diet pills, detox programmes and unsupported health claims.
Some categories are restricted rather than banned, and the restrictions are where Indian advertisers get caught:
- Financial services are allowed case by case, for approved advertisers in the US only. Outside the US they're generally prohibited. In practice that covers most lending, payments, investment and insurance brands targeting India.
- Health services follow the same rule: US only, case by case, generally prohibited elsewhere. Online pharmacies and clinics outside the US are effectively out.
- Legal services are US only, and only where the advertiser is licensed.
Separately, ads never appear next to conversations about personal health, mental health or other sensitive topics, whoever the advertiser is. Read the current policy page before you build anything, because OpenAI updates it regularly.
Can you buy them from India?
Yes. Self-serve Ads Manager opened to advertisers in India on 31 August 2026, and OpenAI's availability page now lists India among more than 60 self-serve countries. The legal entity that advertises and gets billed must be based in one of those countries. Sign-up is at ads.openai.com. Ads are rolling out to users gradually, so check in Ads Manager which audiences you can actually reach before you commit a budget.
Measuring results without fooling yourself
Ads Manager reports impressions, clicks, spend, CTR, average CPC, average CPM and conversions. Two details save you from bad calls in the first week:
- Impressions, clicks and CTR refresh about every 15 minutes, but spend can lag 7 to 8 hours. A spend of zero on day one doesn't mean nothing has been charged.
- You can add your own tracking parameters, like UTMs, to the landing page URL, and they stay on the click.
Use that second point. ChatGPT already sends organic visitors with utm_source=chatgpt.com, which we covered in how to track AI search traffic in GA4. Add your own parameters to ad URLs, for example utm_medium=cpc&utm_campaign=chatgpt-ads, and check in GA4 which values actually arrive on paid clicks. Otherwise paid and organic ChatGPT visits blur together, and you can't tell whether the ads are doing anything your organic visibility wasn't already doing.
A first-campaign checklist
1. Check you're eligible
Confirm your country is on the self-serve list and your category isn't prohibited or US-only. If you're in fintech or health in India, stop here and put the budget into organic AI visibility instead.
2. Pick one situation and one landing page
Start with a single conversation type that leads directly to your product. Send traffic to the page that answers that exact situation, not your home page.
3. Write context hints as situations
Three to six per ad group, each describing what someone is trying to do. Leave out your brand name and generic category terms.
4. Write copy that adds to the answer
Assume they've just read a good answer. Your title and description need a reason to click that the answer didn't give: a price, a free tool, a template, a specific outcome.
5. Start on Clicks with OpenAI's suggested bid
Set a daily budget you're comfortable losing while you learn. Move to the Conversions objective once your conversion event is set up and recording correctly.
6. Tag and set up conversions before launch
Add your UTMs, set up the conversion event in Ads Manager, and confirm a test visit shows up in GA4 before any spend goes out.
7. Judge it after a week, not a day
Spend data lags, and early auctions are noisy. Compare cost per conversion against your other paid channels, and check whether ChatGPT ad visitors behave differently on your site from organic ChatGPT visitors.
Should you test it now?
It's worth testing now if your product is something people ask ChatGPT for help choosing or using, your category is allowed outside the US, and you can measure conversions properly.
Wait if you can't track conversions yet, if your category is restricted, or if your buyers mostly use ChatGPT through paid or business plans that don't see ads. In those cases, organic visibility in ChatGPT's answers does more for you, and it costs nothing per click. We explain how that works in what generative engine optimization is.
If you'd rather have the campaign set up and run for you, that's what our ChatGPT ads management service covers, from eligibility checks to context hints and measurement.