ChatGPT Ads - How Advertising in ChatGPT Works in Poland

As of September 7, 2026. Dashboard screenshots from the same day. Ads Manager is in beta and has changed several times this year, so before launching a campaign, check OpenAI's current documentation.

Ads in ChatGPT are a separate sponsored card beneath the model's response, clearly marked as an ad. Poland has been on the list of markets where you can purchase them self-serve via the Ads Manager dashboard since August 31, 2026.

We went through the dashboard ourselves, up to the point just before launching delivery. Below is what is actually in there, how it differs from Google Ads, and what cannot be purchased in a campaign targeting Poland.

How ads in ChatGPT work and where exactly they appear

The sponsored card appears only after a fully generated response. Never within its text, never while it is generating. The ad placement is visually separated from the organic response and marked as sponsored, and either a single card or several may appear beneath one response (Ads in ChatGPT, OpenAI Help Center).

The response mechanism itself remains untouched. An ad does not alter the model's training, the prompt, or the response logic; it simply attaches as a separate UI component. Here lies the biggest difference compared to Google results, where sponsored links are mixed with organic ones within a single module: a ChatGPT user first gets the complete response and only sees the card afterward.

The timeline is brief and dense. Testing began on February 9, 2026, in the US (Testing ads in ChatGPT). Canada, Australia, and New Zealand were added in March, followed by the UK, Mexico, Brazil, Japan, and South Korea in May. On August 18, 2026, OpenAI announced an expansion into 31 European markets, its largest to date (ChatGPT Ads expands across Europe). Those 31 represent an expansion, not the total count: along with the nine earlier markets, this brought the total to about forty countries, and the Ads Manager availability list now has over fifty entries, including Poland (Ads Manager Availability).

The entry requirement applies to the entity, not the audience. The business paying for the campaign must be based in an eligible country, meaning a Polish company buys ad delivery directly, without an intermediary.

Channel scale: what OpenAI shares and what it keeps private

Colin Fleming, OpenAI's Chief Marketing Officer, reported one billion weekly active users at the end of August 2026, adding that roughly 20 percent of them show commercial intent (Adweek). The second figure matters more than the first because it is what defines the market size for an advertiser, not raw reach alone.

On the same day, OpenAI reported that ads had reached an annualized revenue run rate of one billion dollars, 203 days after testing began. This is a twelve-month forecast calculated from current sales, not realized revenue, and it comes from the seller. In our view, treat it as an indicator of momentum rather than proof of performance: annualized figures for a nascent channel can rise and fall within the same quarter.

Who will see your ad: tiers, age, ad-free modes

Segmentation in ChatGPT Ads is simpler than in Google Ads or Meta Ads. There is no demographic or behavioral targeting. Instead, segmentation is based on the subscription tier, account age, and user-configured settings.

• Ad-supported plans: Free and Go. Sponsored cards appear only on these tiers. Plus, Pro, Business, Enterprise, and Edu are ad-free (Ads in ChatGPT).
• Logging in changes inventory, but does not dictate delivery. Ads also appear in logged-out sessions, but only those classified by OpenAI as suitable for general audiences. Age-restricted creatives are served exclusively to eligible logged-in accounts.
• Adults only. OpenAI does not display ads to accounts identified as belonging to users under 18, based on account data and behavioral signals.
• Ad-free temporary chat. Sessions designated as temporary are excluded from ad delivery regardless of topic or plan.
• Three different opt-out paths. Upgrading to an ad-free plan, using Ads-Free mode on the Free plan, or simply disabling personalization. Under the third option, ads remain, they are just less tailored - and in Europe, they were not personalized anyway.

Ads-Free mode comes at a cost: fewer messages and the loss of certain tools, including image generation and deep research. Opting out of ads cuts off deep research in particular - the very work mode closest to a purchasing decision - and that alone narrows the pool of people who use it.

Context Hints versus keywords: how it differs from Google Ads

ChatGPT Ads does not run keyword auctions. Advertisers provide Context Hints, which describe conversations, topics, and queries where their offer is relevant. OpenAI explicitly notes that these are not exact-match terms and do not guarantee delivery. In the dashboard, this field is marked as optional.

The system evaluates the topic and intent of the entire conversation, rather than a single entered query. The focus shifts from lexical matching to situational understanding.

The row regarding control over terms captures the essence of the shift. In Google Ads, you build an account around keyword lists and must anticipate query variations. Here, that effort shifts to describing the scenario where someone is discussing your topic with the model.

This changes how campaigns are built. Instead of pulling terms from a keyword research tool, you need to understand what users write to the model when seeking help in your domain: not "seo services pricing," but rather the underlying problem they present. In our view, this is the only part of the system where Google Ads experience does not directly transfer, and the only area where it pays to spend time on the craft rather than copying and pasting existing lists.

What cannot be purchased in Poland and Europe

ChatGPT Ads in Poland and Europe face strict regulatory limitations. Personalization is unavailable in the European Economic Area and Switzerland, which eliminates remarketing and Custom Audiences. OpenAI states this in its developer documentation: "Custom audiences are not supported for campaigns targeting the European Economic Area (EEA) or Switzerland, where personalized ads are not yet available" (Custom Audiences, OpenAI Developers). The dashboard displays this exact caveat when selecting location targeting, as shown in the screenshot below.

Privacy here is a systemic requirement, not just a policy statement. OpenAI does not share conversation content, memory data, names, email addresses, or sensitive data with advertisers. The dashboard provides aggregate reporting: impressions, clicks, spend, CTR, average CPC, average CPM, and conversions. Advertising data is deleted after at most 30 days. Optimization therefore relies on aggregates, which rules out several tactics carried over directly from Meta Ads.

Topics you cannot promote on ChatGPT

OpenAI restricts ad delivery alongside content related to physical health, mental health, and politics. Political advertising is banned entirely. Budget makes no difference here.

• Politics - parties, candidates, election campaigns, and politically persuasive content are ineligible for delivery in any context.
• Personal and mental health - ads do not appear alongside conversations about medical conditions, diagnoses, therapy, or mental health, even if the topic was initiated by the user.

The system operates based on the recognized topic of the conversation, not the submitted phrase. If the system classifies a conversation as touching on one of these categories, no card will be generated.

A note on the scope of this restriction, as it is often misunderstood: OpenAI restricts conversation context, not the advertiser's industry. A pharmaceutical manufacturer is not automatically barred from the platform. What is excluded are conversations about health, which, for such an offering, effectively eliminates most contexts where an ad would make sense. In our view, for pharma, telehealth, and health insurance, it is more realistic to assume this paid channel does not exist - this is our editorial conclusion from reviewing the policies, not an official list provided by OpenAI.

Inside Ads Manager: what the dashboard looks like

The dashboard is simpler than Google Ads and does not stray far from what we know from other platforms. The builder has three steps: campaign, ad group with creative, and review. Below are four screenshots from September 7, 2026, from an account billed in Polish zloty (PLN).

Account: verification and irreversible settings

The self-service dashboard runs at ads.openai.com. Onboarding has five steps: account creation, verification, business details, billing, and team invitation (Ads Manager Beta Account Setup). Identity verification goes through Persona, and OpenAI checks along the way whether your offering complies with ad policies. Without a billing profile and a payment card, no creative will go live.

One thing to remember before clicking "next": Country, currency, and time zone are set during account creation and cannot be changed later. For an agency setting up client accounts, this is the most critical sentence in the entire guide, because a mistake means a brand-new account, not editing settings.

Campaign objective and location

First step of the builder: campaign name, objective from three options, included and excluded locations.

The campaign objective has three options: Reach, Clicks, and Conversions. This choice dictates the rest of the configuration, as it determines the billing model and what the system optimizes delivery for.

Locations are set on two fronts. You separately specify the ones your campaign should target and the ones you want to exclude. The dashboard notes that a user in an excluded location will not see the campaign even if they match an included location, meaning exclusions take precedence. Targeting levels include country, region, and DMA (Campaign Targeting).

Budget, platforms, and automatic text rewriting

Campaign level: platforms, daily budget, and the "Text customization" toggle, enabled by default.

Three things on this screen are worth knowing before estimating campaign costs.

The minimum daily budget for an account billed in PLN is 65 PLN. The dashboard will not accept a lower amount and tells you this via validation, not just a tooltip. That is around 2,000 PLN per month as the barrier to entry for a single campaign on this channel, and it is a hard number straight from the interface, not an estimate.

The budget is an average amount, not a rigid cap. For any entered daily amount, the dashboard specifies its own spend caps: in our test, twice the daily amount on any single day and seven times over seven days. When planning spend, calculate using the weekly upper cap, not just daily multiplication.

The third detail is the most interesting and easiest to overlook. The Text customization toggle is enabled by default, and its description states that AI automatically generates personalized and translated versions of headlines and descriptions, which may be displayed without advertiser approval each time. The creative you approved is therefore not necessarily the creative the user will see, unless you turn this toggle off. In our opinion, it is worth disabling for your first campaign for two reasons: there is no other way to assess what actually works, and for a brand with an established tone of voice, automatic headline rewriting is an unrequested risk.

Eligible platforms are set separately, with all selected by default.

Bidding strategy and contextual hints

Ad group level. "Maximize results" is selected by default, and contextual hints are marked as optional.

Separate from the campaign objective, you configure the bidding strategy, and this is where the catch lies. Maximize results hands bids over to the system, which adjusts them toward your campaign goal within your budget. Manual: max bid leaves the bid up to you. The automated option is checked by default, so advertisers unaware of it will have it turned on (Maximize Results Bid Strategy).

OpenAI does not guarantee a target CPA, CPC, or ROAS with it, only result volume. It is worth running your first campaign manually specifically to establish your own benchmark for what the automation delivers later. As a starting point for click-focused campaigns, OpenAI suggests 3-5 dollars per click. This is the only bid figure published by the provider, and it is a reference point, not a market price.

Landing page query parameters are entered in the same place. Supported macros include {campaign_id}, {ad_group_id}, {ad_id}, {ad_account_id}, and {oppref}, allowing you to assemble analytics attribution without manually tracking UTMs on every single creative.

Creative: character limits and images

Creative form. Title up to 50 characters, description up to 100, square image with a minimum of 256 by 256 pixels.

The limits are tight: title up to 50 characters, description up to 100 characters. For comparison, Google Ads offers three times more space for search ad copy alone. The image must be a PNG or JPG file, and the dashboard recommends a square image of at least 256 by 256 pixels.

The card also displays the advertiser name along with a favicon, meaning the creative consists of five elements, four of which you control (Ads in ChatGPT: The Basics).

Pixel, Conversions API, and UTMs

Measurement relies on two mechanisms. The Measurement Pixel is a JavaScript snippet on your landing page that sends standard and custom conversion events. The Conversions API sends the same events from your backend, making tracking less dependent on browser cookies (Ads, OpenAI Developers). A separate Advertiser API is available for managing campaigns and querying performance data.

Online stores have an additional option: Ads Manager can build campaigns from a product catalog feed. For e-commerce with a wide inventory, this is the only operationally viable path.

Paid ads are not your entire presence on ChatGPT

Sponsored cards reach only logged-in users on Free and Go tiers, as well as logged-out sessions from limited inventory. Beyond that scope lies what the model says about your brand in organic responses, and that cannot be purchased inside Ads Manager.

A few factors practically determine whether the model has material to quote you from:

• Answers in a single, self-contained snippet. Models represent content as semantic vectors, not as strings of keywords. If an answer to a specific question is scattered across four paragraphs, none of them makes a good candidate for citation. We broke down this mechanism in depth in our article on how AI algorithms work.
• Rendering for agentic mode. Atlas, OpenAI's standalone browser, was discontinued on August 9, 2026, and web agent workflows moved to the ChatGPT desktop app and Chrome extension (OpenAI Help Center). The underlying principle remains identical: the agent reads the page during the session, so slow client-side rendering and a lack of clear heading hierarchy will still work against you.
• Structured data. Schema.org paired with clearly attributed authorship, publication dates, and factual sources increases the likelihood that the model will turn to your page rather than someone else's.
• A shared event framework. If you run campaigns in Ads Manager while simultaneously building organic visibility, measure both channels using the same conversion events. Otherwise, you are comparing reports rather than actual results.

We covered summaries generated in Google Search separately in our piece on Google AI Overviews. If you are looking for an organized, end-to-end framework rather than isolated tweaks, this is exactly what we handle as part of our AI search engine visibility services.

Frequently Asked Questions About Ads in ChatGPT

Do ads in ChatGPT affect the model's responses?

No. The card is a separate interface component placed below the completed response, and it has no access to its content or the logic behind its generation.

Can I advertise on ChatGPT if I have a Polish company?

Yes. The requirement applies to the entity, not the audience: the company that pays and is billed must be based in a country on the availability list, and Poland is on it.

How much does it cost?

For an account in Polish zlotys, the minimum daily budget is 65 PLN. Other than that, OpenAI does not publish a rate card because the cost is determined by auction.

How do Context Hints differ in practice from keywords?

You cannot generate a search terms report or add negative keywords like in Google Ads. Instead of a list of phrases, you prepare a description of a situation in which someone is talking to the model about your topic.

Why doesn't remarketing work in my campaign targeting Poland?

Because ad personalization is not available in the EEA and Switzerland, and Custom Audiences are disabled in such campaigns. This is a built-in platform limitation, not a configuration error.

Can AI change my ad creative without asking?

Yes, if you leave "Text customization" enabled. The dashboard will then automatically generate personalized and translated versions of the headline and description without requiring individual approval each time.

How do users turn off ads?

Three ways: an ad-free plan, the Ads-Free mode on the Free plan, or simply turning off personalization. Only the first two remove ads, and the second option costs you message limits and access to some tools.

Can logged-out users see ads?

Yes, but from a limited inventory: exclusively creatives suitable for all audiences.

Sources

• ****Ads in ChatGPT**** - OpenAI Help Center
• ****Ads in ChatGPT:** The Basics** - OpenAI Help Center
• ****Ads Manager Availability**** - OpenAI Help Center
• ****Ads Manager Beta Account Setup**** - OpenAI Help Center
• ****Testing ads in ChatGPT**** - OpenAI, February 9, 2026
• ****ChatGPT Ads expands across Europe**** - OpenAI, August 18, 2026
• ****ChatGPT Surpasses $1 Billion in Annualized Ads Revenue**** - Adweek, August 31, 2026
• Ads Manager dashboard, account billed in PLN, as of September 7, 2026. Own screenshots.