ChatGPT Ads in Poland: advertising is only the beginning.
Businesses need to think about AI visibility

Until recently, the question was: will ChatGPT have ads? Today, a better question is: how large an advertising and search channel could it become?

Since 31 August 2026, the self-service ChatGPT Ads system has also been available to advertisers in Europe, including Poland. But that is only the most visible part of a bigger shift. [3, 4]

In just a few years, language models have gone from a technological curiosity to tools used by hundreds of millions of people to find information, compare products, plan purchases, choose services and make business decisions.

In August 2026, OpenAI reported more than a billion weekly active ChatGPT users. Meanwhile, ChatGPT's advertising business reached roughly US$1 billion in annualised revenue run rate in fewer than 200 days and attracted tens of thousands of advertisers. [4]

This does not mean the end of Google, or a revolution after which traditional SEO ceases to exist. Quite the opposite.

SEO is not disappearing. It is expanding into more places where algorithms need to understand what a business does, what it offers and why it should be treated as a credible source.

From experiment to advertising platform in a few months

OpenAI first officially announced plans to test advertising in ChatGPT on 16 January 2026. Initially, ads were to be shown to users of the Free plan and the lower-cost Go plan. [1]

The test actually launched in the United States on 9 February 2026. From the outset, ads were separated from the model's answer and labelled as sponsored. OpenAI also says ads do not influence ChatGPT's answers and advertisers do not receive access to users' private conversations. [2]

The platform developed quickly. After around six weeks of the pilot, Reuters, citing an OpenAI spokesperson, reported that annualised advertising revenue had passed US$100 million. The system then had more than 600 advertisers. Around 85% of Free and Go users in the US were eligible for ads, but fewer than 20% of them saw ads on any given day. [5]

Over the following months, ads expanded into more markets. On 18 August, OpenAI announced ChatGPT Ads' arrival in 31 European countries, making its self-service Ads Manager available there on 31 August. [2, 3]

On the same day, the company announced that, in fewer than 200 days since launch, ChatGPT Ads had reached a US$1 billion annualised revenue run rate, tens of thousands of advertisers were using the system, and ChatGPT had more than a billion weekly active users. [4]

The word “annualised” matters. It does not mean OpenAI had already booked a billion dollars in advertising receipts. Run rate describes the current pace of the business extrapolated over a full year. Even so, the growth is difficult to ignore.

A billion weekly users changes the perspective

In September 2025, when launching Instant Checkout, OpenAI reported more than 700 million people using ChatGPT each week. Less than a year later, the company was reporting more than a billion weekly active users. [12, 4]

The picture is even more interesting from a Polish perspective. According to a Gemius/PBI report, ChatGPT recorded 9.3 million real users in Poland in June 2025, reaching almost a third of Polish internet users. Gemius notes that measurement of the ChatGPT mobile app was added to Mediapanel on 26 February 2025, so older data series are not fully comparable. [8]

In March 2026, around 13 million people in Poland used tools based on large language models each month, and 9.2 million used them at least once a week. Year on year, monthly reach grew by 71%, weekly reach by 106%, and average monthly usage time by 80%. ChatGPT alone reached around 35% of Polish internet users each month. [9]

At that scale, it is difficult to treat AI solely as a tool for developers, students or technology enthusiasts.

An ad in a conversation is not the same as an ad beside a keyword

The most interesting part of ChatGPT Ads is not the presence of ads itself. We have enough advertising online already. It is the context in which an ad can appear.

For years, the traditional search-advertising model depended primarily on a user's query. Someone types:

“WordPress hosting”

and the advertising system tries to match an ad to the phrase, its meaning and additional signals.

A conversation with an AI model can look very different:

“I run a small WooCommerce shop. I get around 30,000 visits a month, and the site slows down badly during campaigns. I'm looking for hosting in Poland for up to around PLN 300 a month, ideally with an easy migration.”

For an advertising system, that provides much richer information about intent.

OpenAI confirms that, when choosing an ad, the system may analyse the current conversation's context and intent, the landing-page content, the ad headline and description, and advertiser-supplied context hints, among other things. Context hints are not exact-match keywords and do not guarantee placement in a particular conversation. [6]

For years, search engines tried to infer the intent hidden behind a few words. In an AI conversation, users often describe their intent precisely themselves.

How much does advertising in ChatGPT cost?

Technically, Ads Manager increasingly resembles established advertising systems. It currently supports CPM and CPC campaigns, as well as conversion-optimised campaigns, including oCPC. Daily and lifetime budgets, geographical targeting, platform selection and custom audiences are available. [7]

In Poland, the minimum daily campaign budget is currently PLN 65. A daily budget is treated as an average over seven days: daily spending may reach twice the set budget, but spending over the relevant seven-day period should not exceed seven times that budget. [7]

For CPC campaigns, OpenAI recommends starting with a maximum bid of around US$3–5 per click. Ads participate in a relevance-weighted second-price auction, so the highest bid alone does not guarantee delivery. [6]

Ads Manager reports impressions, clicks, spending, CTR, average CPC, average CPM and conversions, among other metrics. According to the documentation, impression, click and CTR data refresh approximately every fifteen minutes, while spending data may lag by seven to eight hours. [6]

This is still a very young system. We do not yet have several years of data and benchmarks comparable to Google Ads or Meta Ads. Claims that ChatGPT Ads will always be cheaper, better or convert more effectively would therefore be premature.

AI increasingly appears before a purchase

Even more interesting developments are happening beyond paid advertising. ChatGPT increasingly helps people choose a product or service.

Users do not need to know which model of equipment they need. They can describe their use case, budget and constraints, then ask AI to build a shortlist. A similar process works when looking for hosting, an accountant, a camera, a phone, a SaaS tool, a website developer or a hotel.

According to Similarweb and Statista's State of Ecommerce 2026 report, around one in four shoppers uses AI before buying. At the same time, direct referral traffic from generative AI currently accounts for only about 0.4% of retail e-commerce visits. [10]

Those figures appear contradictory. The difference lies in AI's place in the buying journey: a system may help create a shortlist, but the user may then type the chosen brand into Google, visit its website directly or buy on a marketplace. Standard analytics may attribute the final visit to another channel, even though the selection process began in ChatGPT or another assistant.

0.4% of the market — but 203% growth

At this point, anyone predicting the immediate end of traditional search engines needs a reality check. According to Similarweb, generative AI currently accounts for around 0.4% of retail e-commerce visits. [10]

That is still very little. The growth rate matters, though: GenAI referrals increased by 203% year on year, faster than the other referral channels analysed in the report. [10]

Combining channels is even more interesting. According to Similarweb, buying journeys involving both AI and Search converted in 23% of cases — 84% more often than journeys using AI alone. [10]

That supports a much more cautious conclusion: at this stage, AI is not replacing search engines. It is beginning to work alongside them and changing what happens before a website visit.

AI traffic may be small, but highly valuable

Adobe's US e-commerce data shows a similar relationship. In July 2026, traffic from AI tools to US online shops increased by 62% year on year; between October 2024 and July 2026, the increase was 1,219%. [11]

The quality of that traffic is even more interesting. According to Adobe, visitors arriving at US shops from AI tools converted around 60% better than non-AI traffic, spent 59% more time on site, had a 33% lower bounce rate and added products to baskets 28% more often. [11]

These figures concern the US market and cannot automatically be applied to Poland or every industry. They do, however, illustrate an interesting mechanism: someone who has already discussed options with AI, compared alternatives and narrowed their choice may reach a website much further down the buying funnel.

ChatGPT also wants to be a place where we buy

On 29 September 2025, OpenAI launched Instant Checkout in the US. Free, Plus and Pro users could buy products from selected Etsy sellers directly within a conversation. The mechanism was based on the Agentic Commerce Protocol, developed jointly with Stripe. [12]

ChatGPT's current shopping features can show products with images, details and seller offers, and for some eligible products and merchants also enable purchases within ChatGPT. [13]

Crucially, OpenAI explicitly separates organic product results from ads. Product results are selected independently and are not advertisements. When choosing sellers, the system may consider availability, price, quality and whether the seller is the manufacturer or primary seller of a product, among other factors. [13]

For Shopify shops, product data can be integrated with ChatGPT through Shopify Catalog. Merchants can also supply OpenAI with a direct product feed to make product information more current and complete. [13]

ChatGPT Ads is only half the story

A website owner may never run a single ChatGPT Ads campaign, yet should still take an interest in ChatGPT's development.

That is because a growing part of finding and choosing may happen before a user visits a company's website. They might ask:

• “Which businesses do X?”
• “Which provider should I choose?”
• “What are good alternatives to product Y?”
• “Where nearby can I order Z?”
• “Which solution makes sense for a small business?”

If AI can answer such questions, a new issue emerges: how does the model know a particular company exists, what it does and whether information about it is current?

SEO for AI? GEO? AEO? The name matters less than the change

Terms such as GEO (Generative Engine Optimisation) and AEO (Answer Engine Optimisation) have emerged in the industry. We can debate whether new acronyms are necessary, but the shift in how people discover information is real.

Google takes a pragmatic approach. Its official 2026 guide explicitly states that established SEO practices remain the foundation of visibility in generative search features too. Google describes AEO and GEO as industry terms, but from its own perspective still treats optimisation for generative search as SEO. [15]

There is no magic “make-this-site-visible-in-ai=true” tag. Google even says that, for its search engine, some fashionable “GEO hacks” can be ignored, including unnecessary AI text files such as llms.txt and artificially generated mentions. Clear technical structure and unique, valuable content should remain the foundation. [15]

The most practical approach is not “abandon SEO and do GEO”. It is “make the company's content and data clear, technically accessible and useful to people, search engines and AI systems alike”.

ChatGPT must also be able to read the website

OpenAI provides its own search crawler, OAI-SearchBot. Website owners who want their content to appear in ChatGPT Search results should ensure that relevant pages are not accidentally blocked from this crawler. [14]

OpenAI also says traffic from ChatGPT results is automatically tagged with utm_source=chatgpt.com, allowing it to be identified in analytics tools such as Google Analytics. [14]

The distinction between crawlers also matters: OAI-SearchBot handles search, while GPTBot can be blocked where a publisher wants to exclude content from potential model training. Search visibility and the use of content for training therefore need not be the same decision. [14]

Traditional SEO becomes the foundation of AI Search

Google describes a similar mechanism for its generative Search features. AI Overviews and AI Mode use Google's core ranking systems and techniques such as retrieval-augmented generation (RAG): fetching current sources from the search index and using them to prepare an answer. [15]

Google also describes query fan-out: a model can break a detailed user question into a series of related queries to gather information about different aspects of a problem. [15]

That has an important consequence for SEO. Users no longer need to type the exact phrase for which we tried to rank a page. They can ask a detailed question, and the system will find the information needed to answer it.

A good website should do more than “contain keywords”. It should contain information the system can use when building an answer.

What does this mean for an ordinary business?

It does not mean every small business should hire a “GEO specialist” tomorrow. Sorting out the basics makes far more sense.

The website should clearly explain who we are, what we do, where we operate, whom we work for, which problems we solve, what products or services we offer, how the options differ and how to contact us.

• Technical accessibility — crawling, indexing, a sitemap, sensible URL structure, internal links and no accidental crawler blocks. [14, 15]
• Clear content semantics — specific names for services, products, locations and parameters rather than marketing slogans alone.
• Content answering real customer questions — guides, comparisons, FAQs, case studies, instructions and expert material offering information beyond the repetitive content found on other websites. [15]
• Structured data and metadata — where applicable, such as organisation, local-business or product data; in e-commerce, up-to-date product information too. [13, 15]
• Current information in external ecosystems — business profiles, directories, marketplaces, Merchant Center and other sources used to verify information. [15]
• Credibility beyond your own domain — real publications, mentions, reviews and industry sources confirming the company's existence and expertise.
• Measuring AI traffic and visibility — including traffic from chatgpt.com and generative-AI reports in Google Search Console. [14, 16]

AI visibility is becoming measurable

Until recently, one of GEO's biggest problems was that people could discuss “AI visibility” at length, but measuring it meaningfully was much harder.

On 3 June 2026, Google announced separate website-visibility reports for generative features in Search Console, covering AI Overviews and AI Mode among others. On 31 August, Google updated the announcement to say the data was available globally to all websites. [16]

The report shows URL impressions in generative-AI features, individual pages, countries, devices and changes over time, among other things. OpenAI, meanwhile, makes ChatGPT clicks identifiable through utm_source=chatgpt.com. [16, 14]

This is an important stage: AI visibility is gradually becoming more than a theoretical discussion, turning into another part of traffic analytics and brand discovery.

Will AI kill Google?

The data does not show that yet. Similarweb actually finds that journeys combining AI and Search convert significantly better than journeys based on AI alone. [10]

Meanwhile, Google is integrating generative AI extensively into its own product. The boundary between “search engine” and “AI” is beginning to blur.

The question “Should we do SEO or GEO?” will probably make less and less sense. A better question is:

“Can the internet clearly understand our company, offer and expertise, whether users look through Google, ChatGPT, Gemini or another system that appears next year?”

You can switch on ads tomorrow. Organic visibility takes time

That is why the arrival of ChatGPT Ads matters even to businesses with no intention of spending any money there.

An advertising campaign can be launched relatively quickly: prepare a landing page, creative assets, campaign structure, a budget and measurement. Organic visibility is different.

A domain's history, publication quality, indexing, external mentions, offer information and organisational credibility develop over months and years. If AI models become another intermediary between users and the internet, this area may matter more to many businesses than the ability to buy an ad.

We do not have to choose between Google and AI

At this stage, the worst conclusion would be: “Google is finished — let's optimise everything for ChatGPT”. The data does not support it.

Direct referrals from AI remain small compared with the overall market. Yet AI is growing at a pace that is hard to miss: LLMs already have 13 million monthly users in Poland, ChatGPT has passed a billion weekly active users globally, and Similarweb reports 203% growth in generative-AI referral traffic to retail e-commerce. [9, 4, 10]

Google is developing AI Mode, AI Overviews and separate visibility reporting for these features in Search Console. OpenAI is developing search, advertising, product feeds and shopping within conversations. [16, 13, 12]

This does not look like a passing trend. It looks more like the next stage in how we find information.

Your website remains the reference point

Paradoxically, AI's development may make having a good website of your own more important, rather than less.

The system needs data it can find, understand, compare with other sources and — for search based on current information — cite or use to construct an answer.

If a company has only a social profile with a few photographs and a phone number, the scope for clearly understanding its offer is limited. A well-built website can clearly describe its activity, location, offer, product specifications, specialisms, experience, completed projects, answers to customer questions and current contact details.

That is exactly the kind of information traditional search engines needed. Generative systems and AI agents are now beginning to need it too. [15]

Conclusion

ChatGPT Ads is an interesting new advertising channel, but the ability to buy an ad is not the most important change.

Far more significant is that users are starting to converse with the internet rather than merely search it. Questions are becoming longer, intentions more precise, and AI increasingly helps narrow choices before a website visit.

Meanwhile, traditional search remains a very important part of that process. SEO is therefore not dying — the environment in which it operates is changing.

A technically sound website, valuable content, clear company information, appropriate indexing and credibility formed the basis of Google visibility. There are growing indications that they will also underpin visibility in a world where AI models increasingly prepare answers to users' questions. [15]

You can buy advertising. Buying credibility and a brand's presence in the information used by search engines and AI systems to build answers is much harder to do in a few days.