A year ago the answer to "how do I advertise in ChatGPT" was: you can't, you can only hope the chatbot mentions your brand on its own. In 2026 the situation changed: OpenAI introduced its own ad format, an ad dashboard and self-serve campaign launch, and since October the channel has been available in dozens of countries, including Kazakhstan.
This article covers how advertising in ChatGPT works, what happened to it in 2026, how to launch it, how much it can cost and what results the first campaigns show. We will also look at who it suits and who it is too early for, and why you cannot "buy a mention in the answer".
What it is and what it looks like
Advertising in ChatGPT is a labelled block marked "Sponsored" that appears under the chatbot's answer. It looks like a card: a headline, a short description, sometimes an image and a button linking to the advertiser's website. The ad is separate from the answer itself, and OpenAI states that it must not influence what ChatGPT actually tells the user.
The fundamental difference from familiar contextual advertising is that a person does not type a short two- or three-word query but holds a conversation. They can describe the task in detail, for example: "I'm choosing a CRM for a small online store, the budget is limited, I need Kaspi integration". That extended context is what is used to decide whether an ad is appropriate under the answer. For a business, it is a chance to appear at the moment a person is already formulating a specific need.
Timeline of 2026
To see how young the channel is, it is enough to look at its history over a single year.
- less than four months passed between the announcement of ads and self-serve campaign launch for advertisers;
- nine months passed before the move into international markets;
- on 21 October 2026 the list of countries was expanded and Kazakhstan was included.
That is fast, and it also means the channel has not settled yet: features are being added, rules are being refined and there is little accumulated statistics.
Is advertising available in Kazakhstan
According to the availability page in OpenAI's ad dashboard (Ads Manager Availability), Kazakhstan is listed as an available country and was added in the expansion of 21 October 2026. This means a Kazakhstani business can, in principle, run campaigns.
There are several nuances worth checking before committing a budget:
- the legal entity that is invoiced must be located in one of the listed countries;
- according to third-party reviews, new accounts may at first be limited to showing ads in the country of registration, so reaching audiences outside Kazakhstan right away may not be possible;
- according to the same reviews, a number of topics - finance, health, legal services - are subject to restrictions or additional checks. If your business is in one of these niches, read the current ad policy in OpenAI's help center before preparing creatives.
One last point: all of this is written as of November 2026. The list of countries, conditions for new accounts and payment methods are exactly what changes fastest.
Who sees the ads and how they are selected
Ads are seen by users of the free Free plan and the low-cost Go plan. Plus, Pro and Business subscribers do not see ads, and neither do users under 18. So the ad audience is not the whole ChatGPT audience but its free part, and this should be factored into planning: paying users, who often make more deliberate purchasing decisions, are out of your reach.
Delivery works as a second-price auction that takes relevance into account. This means the highest bid alone does not win: the ad must also match the context of the conversation. And you usually pay not your maximum bid but a price slightly above the nearest competitor's bid.
Another important feature: there are no keywords in the usual sense. Instead of exact match with a query, the advertiser provides context hints - what the conversation should be about for the ad to be appropriate. The system itself decides whether a particular dialogue fits. On the one hand this widens reach; on the other it reduces predictability: you cannot guarantee that the ad will appear under exactly the answer you had in mind.
How to launch ads: 8 steps
Below is the general order of actions. The dashboard interface keeps changing in beta, so button names may differ.
- Check availability. Open the ad dashboard's availability page and make sure the country of your legal entity is on the list.
- Register in the dashboard. Ads Manager is at ads.openai.com. You will need an account and your organization's details.
- Set up payment. According to available information, billing is postpaid: impressions first, then the charge. Check in the dashboard which payment methods are available for your legal entity.
- Define the campaign goal. Optimization for impressions (CPM), clicks (CPC) and conversions is available. For a first test it is usually simpler to start with clicks.
- Set context hints. Describe the topics and situations in which it is appropriate for your product to appear. Phrase them as customer tasks, not as a set of keywords.
- Prepare the ad. A headline, a short description, a link to a page. Text length limits are cited differently in different sources, so go by what the dashboard itself shows.
- Set up measurement. Add UTM tags to the link and, if you need conversion optimization, install the pixel or connect the Conversions API.
- Launch and observe. The first days are a learning period: do not change parameters every few hours, let the campaign accumulate at least minimal statistics.
How much it costs
The main warning here: there are no reliable market benchmarks yet, and OpenAI itself admits this - according to the company, benchmarks for the channel do not exist yet. All that is available are individual observations from early advertisers and reviews.
According to this data, the cost per click in different campaigns ranged from roughly $3 to $22, so the spread is very wide. In its recommendations OpenAI advises setting a maximum click bid of around $3-5 to start. One third-party source mentions a minimum daily budget of about $25, but this is not official data, and the exact value should be checked in the dashboard when creating a campaign.
A practical conclusion follows for a Kazakhstani business. A click costing several dollars is considerably more expensive than in the usual channels for the local market, so it makes sense to test the channel where customer value is high: complex services, B2B, expensive goods, development and implementation. For goods with an order value of a few thousand tenge, such a click cost will most likely not pay off.
How to measure the result
According to available information, the pixel and Conversions API work only for websites. This means that for a business whose main sales channel is a messenger, phone call or app, measurement will be incomplete and will have to be supplemented manually.
The minimum worth setting up in any case is UTM tags on the ad link. They let you see visits from ChatGPT in analytics as a separate source rather than mixed with other traffic. Then you need to follow the path further: how many of these visits turned into leads and how many leads into paid orders. For this, leads must reach the CRM together with their source, otherwise you will see only clicks and will not understand whether they paid off. If you have had cases where leads get lost between the website and the CRM, deal with that first - otherwise any ad statistics will be distorted.
Ask a question or leave a request
We will build a fast landing page for your offer and set up UTM tags and lead passing to the CRM with the source, so you can see the real cost per customer.
By submitting the form, you agree to the processing of personal data.
What results the ads show
Let's start with an honest limitation: independent data is scarce and the channel is new. OpenAI itself states that there are no unified benchmarks yet, and this should be kept in mind when reading any figures.
According to analysts at Graphite and Similarweb, the average CTR in the US was about 0.73%. Across the nine markets studied, values ranged from 0.60% to 1.91%. For comparison: this is on par with ordinary search and display formats, so there is no sensation in click-through rate, but no failure either.
Feedback from early advertisers is contradictory. Some note that traffic is small but leads are of higher quality, because the person arrives with a clearly formulated task. Others complain about expensive clicks and modest volume. One review mentions that about 14% of the ads shown were only loosely related to the topic of the conversation. That could be a shortcoming of context targeting or a feature of how the system works at an early stage.
The overall conclusion is this: today it is a channel for tests, not for large investments. It may prove profitable in specific niches, but this can only be confirmed with your own numbers.
Who it suits and who it's too early for
It makes sense to try the channel if your customer value is high and a single order pays back dozens of clicks. This includes website and app development, CRM and 1C implementation, consulting, educational programs and expensive equipment. Another good sign is a product that people look for through questions and comparisons rather than by name: "how does X differ from Y", "how to choose", "what is better for my case".
It is too early for businesses with a small order value and a margin that does not allow paying several dollars per click, as well as companies without a website that has proper analytics. Without measurement you will not know whether the ads work and will spend money blindly. Finally, if your niche is finance, health or legal services, check the restrictions before planning a campaign.
How to run the first test
It makes sense to keep the first test small and clearly limited in time. Decide in advance how much you are ready to spend - for example, an amount you won't mind losing - and for how long. Choose one topic, one offer and one landing page so that the result cannot be explained by ten reasons at once.
The landing page matters more than the ad itself: the person came from a conversation with a chatbot with a specific task and wants to see the answer to it right away. Do not send them to the home page - build a separate page for the ad's topic with a clear next step, such as a request form or a cost estimate. If the page is slow, it will cancel out any optimization in the dashboard, so it is worth checking load speed in advance.
After the test ends, look not at clicks but at how many leads were received, how many of them reached payment and what the final cost per customer is. Only after that can you decide whether to continue, change the approach or drop the channel.
Can you buy a mention in the answer itself
No. Paid ads are shown in a separate labelled block and are not embedded in the text of the answer. You cannot buy a place in ChatGPT's recommendation itself, and OpenAI states that ads do not influence the content of answers.
If you want the chatbot to mention your brand on its own, that is a separate task, solved not by advertising but by reputation and content: a clear description of services on the website, detailed expert articles, mentions in authoritative sources, reviews. It is long work without a guaranteed result, but it requires no ad budget and accumulates over time. Incidentally, the quality of the website itself belongs here too: a technically tidy, fast website with a clear structure is better perceived by both search engines and chatbots.
Risks and limitations
The first risk is unstable conditions. The channel is in beta, rules and prices change, and what works today may change in a month. The second is the lack of benchmarks: there is nothing to compare your results against, and it is unclear what counts as a good figure. The third is a limited audience: subscribers of paid plans do not see ads. The fourth is incomplete measurement if your sales do not go through a website.
Add to this dependence on a single platform: if you invested only in this channel, any rule change hits your leads directly. So it is sensible to treat it as one experiment in the overall marketing portfolio, not as a replacement for search, social media and your own website.
Pre-launch checklist
- The country of the legal entity is on the ad dashboard's availability list.
- The niche does not fall under ad policy restrictions.
- A test budget and test duration are defined.
- There is a separate landing page for the ad's topic.
- The page loads quickly and opens correctly on a phone.
- UTM tags are set up on the link.
- Leads reach the CRM with the source specified.
- It is decided which metric counts as success (cost per lead, cost per customer).
- Limits, fees and payment methods are checked in the dashboard, not only from articles.