OpenAI opened ChatGPT ads to users in Turkey on 17 September 2026. Ads appear only on the Free and Go plans; Plus, Pro, Business, Enterprise and Education users see none. The placement is fixed: the ad sits beneath the answer, carries a "sponsored" label and lives in an area visually separate from the answer itself. The self-serve ads panel for Turkey is open at ads.openai.com/tr-TR.
Turkey is the latest stop on a long timetable. OpenAI began testing in the United States on 9 February 2026; Canada, Australia and New Zealand followed on 26 March, then the United Kingdom, Mexico, Brazil, Japan and South Korea on 11 August. The 18 August announcement added 31 European countries and the self-serve Ads Manager opened on 31 August; on 4 September OpenAI announced expansion to selected countries across Europe, India, the Middle East and North Africa along with new pieces of the ad stack. What arrived in Turkey is the product of a seven-month test, not an experiment.
This is a decision piece, not a news report. The thesis fits one sentence: ads go beneath the answer, GEO goes inside it. Neither replaces the other; but for a business with a budget, "being visible in an AI answer engine" has become something you can buy for the first time. We are writing for three readers: the commerce founder wondering whether to move budget away from Google Ads, the industrial transformation manager asking whether their customer decides inside ChatGPT at all, and the marketing manager who has to build the measurement. The second half of the article takes the same question through the SME, the exporter and the large company in turn — because a request is worth the most at the moment it is wanted and wondered about, and those three catch that moment from different places.
How do ChatGPT ads work?
The ad appears once ChatGPT's answer has finished, in a separate area beneath it, and carries a "sponsored" label. It does not blend into the answer text: OpenAI's commitment is that ads do not influence the answers ChatGPT gives. Today's format is an image, a headline and a destination link; a carousel format fed from a product feed also exists. Video is not yet available, according to the Turkish press.
Matching runs on context, not keywords. According to OpenAI's own text, four signals feed ad selection: the topic of the conversation, past conversations, previous ad interactions and memory data. The advertiser sees none of these signals; what reaches them is aggregate metrics such as impressions and clicks. No individual conversation, history or identity data is shared, and OpenAI writes that conversations are not shared with advertisers and customer data is not sold.
On the user's side there are four controls: dismissing an ad, asking "why this ad", deleting ad data and switching personalisation off altogether. There is also an exclusion rule: no ads appear next to sensitive topics such as health, mental health and politics. For a brand selling health products that sentence is a boundary, and the ad team needs to know it from day one.
Who sees the ads, and who does not?
Only users on the Free and Go plans see ads. Plus, Pro, Business, Enterprise and Education stay ad-free, and accounts belonging to users under 18, or estimated to be minors, are not shown ads either.
The split tells a marketing plan three things. First, Turkey is a market dominated by the free plan, so the reach is wide. Second, you cannot reach the paying user through this channel — the executive on a corporate account, the team on a Business plan; that reader sees no ad beneath the answer, only sources inside it. Third, the age filter narrows the channel from the start for brands selling to a young audience.
So "being visible in ChatGPT" is not one thing. Visibility to the free user can be bought; visibility to the paying user is possible only as a source of the answer. That distinction is the spine of everything that follows.
How do you place an ad, and what does it cost?
Ads are placed through the self-serve Ads Manager; for Turkey the address is ads.openai.com/tr-TR. A campaign objective is one of three: reach, clicks or conversions. Pricing runs on CPM (cost per thousand impressions) and CPC (cost per click), with conversion-optimised bidding available for conversion campaigns.
Targeting has two layers: country and device selection, plus custom audiences — you upload a hashed email or phone list to reach your own customers or prospects. Those are the layers OpenAI lists; interest targeting is not among them, and the conversation's context does the matching. For measurement, OpenAI offers its Pixel, a Conversions API and third-party measurement integrations.
On cost, the only figure OpenAI itself has published is an incentive: new advertisers who spend 500 dollars receive 500 dollars in credit. According to the Turkish press the minimum daily budget is 25 dollars and there is no personalised remarketing at launch; neither point appears in OpenAI's own text, so we report them on the press's authority. OpenAI said in August that "tens of thousands of marketers" had already advertised — the channel was not empty before Turkey opened.
The self-serve flow runs in this order: open an account, install measurement (the Pixel or the Conversions API), choose a campaign objective, set country and device, upload a custom audience if you have one, enter the image, headline and link, then define budget and bid. The costliest step is not the first but the second: an account that leaves measurement for later cannot read its first four weeks.
Do ads influence ChatGPT's answer?
According to OpenAI, no: ads do not influence the answer, they sit in a separate area, and advertisers have no way to shape what the model says. That commitment is the product's most important sentence, because the channel's credibility rests on it; the day users believe the answer has been bought, neither the answer nor the ad works.
One contradiction is worth recording. Webtekno wrote that conversation history is not used in ad selection; OpenAI's own text counts past conversations and memory data among the matching signals. We go with the primary source: past conversations influence the ad, not the answer.
Not influencing the answer is not the same as not influencing the screen. When the user finishes reading and scrolls down, the first thing they now see is an ad. The brand named inside the answer and the brand advertising beneath it share the same screen; one sells trust, the other sells reach. The ad does not change the answer, but it changes the share of the screen — we will come back to that sentence in the GEO section.
How is this different from Google Ads?
The biggest difference is the intent signal: Google Ads bids on a keyword, a ChatGPT ad matches the context of a conversation. The user does not type "cart abandonment rate"; they explain that "customers give up at the payment step, what should I do", and the ad appears beside that account.
The second difference is funnel position. A search ad usually sits close to the moment of decision; a ChatGPT conversation usually happens while the user is learning, comparing and narrowing options. The third is measurement maturity: the conversion tracking, attribution models and in-channel experiment tools Google Ads has accumulated over the years are still being built here; the Pixel and the Conversions API exist, but according to the press there is no remarketing layer at launch.
- Intent signal — Google Ads: a keyword, the user types what they are looking for. ChatGPT ads: conversation context, the user describes their problem.
- Funnel position — Google Ads: close to the decision, purchase intent already formed. ChatGPT ads: the research and comparison stage, intent still taking shape.
- Measurement and remarketing — Google Ads: mature conversion tracking and retargeting. ChatGPT ads: the OpenAI Pixel and Conversions API exist; retargeting, according to the press, is absent at launch.
The three differences point to one conclusion: ChatGPT ads come on top of Google Ads, not in place of it. The search ad catches the moment of decision; the ChatGPT ad catches the conversation in which the decision matures. The right frame is not moving budget but extending the funnel one level up — the channel mix in our performance marketing service is built on exactly that logic.
Why is the moment of demand so valuable?
A request is worth the most at the moment it is wanted and wondered about; afterwards it only gets cheaper. Suppose a user types this into ChatGPT: "I run a forty-person production plant, we still track stock by hand, where do I start." That sentence is not a keyword; it is a request carrying constraint, scale, urgency and context in one place — the first sentence of a buyer who has not yet gone to a search engine, spoken to anyone or written a shortlist.
Timing is what makes it valuable. The word typed into Google comes several steps after the sentence formed in a conversation: by then the user has learnt what they are looking for, narrowed the options and switched to the language of comparison. In the conversation, the shortlist is being written right now. The brand named inside the answer enters that list as a source; the brand shown beneath it is added as a visible option. Both belong to the same moment, and that moment is the earliest point in the funnel that can be bought.
The moment also carries a responsibility. The user writes that sentence inside a relationship of trust; if the ad beneath the answer shouts, it spends the trust the answer has just built. What context matching rewards is not a loud promise but a single offer that fits the stated problem exactly. This is where the article's thesis becomes concrete: at the moment of demand the ideal position is both inside the answer and beneath it. The inside is earned with content, the space beneath is rented with budget; the brand that holds both at once is the hand writing the shortlist.
How does this relate to GEO: paid or organic?
Ads go beneath the answer, GEO goes inside it. What you buy with an ad is the space next to the answer; what you earn with organic visibility is being named as a source or a recommendation in the answer itself. The economics differ: the ad ends when the budget ends, the citation stays.
GEO — generative engine optimization — is the work of getting answer engines such as ChatGPT, Gemini and Perplexity to use you as a source. Our guide to standing out in AI search explains the method; the Google AI Overviews article covers the Google side of the same logic and the llms.txt article the technical entry point. The common thread: getting inside the answer is done with content, and it cannot be bought.
Between the two kinds of visibility sits a third: the silent citation. Your brand is not named in the answer text, but your page stands in the source list. The reader may scroll past it; even so, your page has entered the model's source pool, and that is the precondition for being named in the next answer. An ad adds nothing to that pool. The paying user, the corporate account, the under-age user and the sensitive topics are all outside the reach of ads — not outside the reach of citations.
Which is why our recommended order does not change: measure first, advertise second. The GEO Visibility Checker measures in a few minutes how readable your site is for answer engines; know where you stand inside the answer before you open an ad budget. To a brand with no organic visibility, an ad sells short-lived reach; to a brand that already has it, the ad hands over the second half of the screen as well. How we run the work of becoming the answer's source is set out in our GEO consulting service.
What does it mean for an SME?
What changes for an SME is the threshold. Being visible in an AI answer engine used to require months of content and authority work; since 17 September a self-serve panel, three campaign objectives, country and device targeting, a press-reported floor of 25 dollars a day and 500 dollars of credit on the first 500 spent let a small budget rent the space beneath the answer. Because the channel is new, competition is also thin.
The SME's real advantage, though, is not budget but sharpness. Context matching rewards a narrowly described offer for a narrowly described problem; a general promise such as "digital agency" fits no sentence in a conversation, while a concrete service or product that solves one problem fits exactly. The large company's broad range scatters in this format; the SME's single product lands in place. For once, this is a channel where an SME long outspent on Google Ads can win on description.
Three cautions apply to SMEs as well. First, measurement: a campaign launched without the Pixel or the Conversions API teaches nothing after four weeks. Second, cannibalisation: set aside a separate test budget instead of cutting the Google Ads spend that already converts. Third, audience: an SME selling to SMEs usually finds its buyer on the free plan, a corporate buyer usually not; who you sell to decides whether the channel fits. An SME running a store that scans its seven critical pages with Diagnoo and removes the friction at checkout before advertising will get more orders from the same budget.
What does it mean for an exporter?
For an exporter the most concrete opportunity is the country list. Ads are live in the United States, Canada, Australia, New Zealand, the United Kingdom, Mexico, Brazil, Japan, South Korea, 31 European countries and selected countries across India, the Middle East and North Africa, and targeting is by country. From the same panel, an offer can be placed beneath the conversation of a buyer in Germany or a distributor in the Gulf. Confirming that the target country is on the list at set-up is all it takes.
The real point is this: the foreign buyer's supplier research increasingly begins as a conversation — "who makes this part in Turkey", "a near-shore alternative supplier for Europe". The Turkish exporter has historically been invisible at that stage; the trade fair, the B2B directory and the intermediary always came later. An ad puts a presence at that stage; GEO puts the supplier inside the answer. The four-language search and answer-engine visibility in the Meccanotecnica Umbra case is the organic form of this; the B2B export platform and supplier portal side of our e-commerce service builds the page that buyer lands on.
Practical rules: the ad's language is the market's language, the destination page opens in the same language, the business cards collected at fairs become the first custom audience as a hashed email list, and each country is measured as its own campaign. The four-week reading rule matters even more in export: buying cycles are long and the first week's data misleads.
What does it mean for a large company?
For a large company the question is not budget but two things: brand safety and where the decision-maker sits. On brand safety the facts are clear — the ad stands apart from the answer under a sponsored label, OpenAI commits that ads do not influence answers, no ads appear next to sensitive topics such as health, mental health and politics, conversations are not shared with advertisers and customer data is not sold. The first questions a corporate compliance team will ask are answered in the product's own design.
The decision-maker question separates two kinds of large company. For a large consumer brand the channel is wide: a market dominated by the free plan, custom audiences built from the CRM list, server-side measurement through the Conversions API and the carousel format; the infrastructure looks designed for the advertiser with data. For a large company selling to corporate buyers the picture flips — its buyer is most likely on a paid or corporate account and never sees the ad. That company's game is inside the answer, as a source.
The strategic point is about defence. The large company already has the content and authority to get inside the answer; for it, the ad is the smaller lever. The risk comes from the other direction: a challenger can rent the space beneath the very answer that names the large brand and take the second half of the screen. So the right posture for a corporate brand is to hold both — a source inside, a visible option beneath. Which conversations to stand beneath is not an ad setting but a brand strategy decision.
Which businesses should try it now, and which should wait?
For e-commerce and consumer brands the answer is: try it, with a small budget and a conversion objective. A market dominated by the free plan means a wide audience; the product-feed carousel and conversion-optimised bidding look designed for this segment. In the GYMWOLVES case, what multiplied sales twelvefold in three months was not a channel but a funnel whose measurement had been repaired; the same rule applies here — measurement first, channel second.
For industry and B2B the answer is: organic first. If your decision-maker researches in ChatGPT, they most likely do so from a corporate or paid account and never see the ad; the only way to reach them is to be the source of the answer. In the Meccanotecnica Umbra case, what multiplied quote requests tenfold was built on an AI technical advisor and search and answer-engine visibility in four languages — not on an ad budget. Ads can come later, as a reach layer in export markets once the organic base is in place.
A business that meets three conditions can start today:
- Conversion measurement works: the Pixel or the Conversions API is installed on the destination page, and the conversion is defined once and clearly.
- There is an offer that fits a conversation: a single product or service that makes sense at the moment the user describes their problem.
- There is budget to keep the channel alive for four weeks: the daily floor is 25 dollars according to the press, but reading the channel needs a meaningful sample.
If one of the three is missing, waiting is not a loss. The channel is new and competition is thin; but a campaign that cannot be measured teaches nothing, and budget that teaches nothing costs more than budget left unspent.
Which 5 steps must the first campaign not skip?
- Measurement first: install the OpenAI Pixel or the Conversions API on the destination page, with one conversion event, tested. A campaign without measurement loses data, not just budget.
- Custom audience: upload existing customer and prospect lists as hashed emails or phone numbers; reading the first round against a familiar audience lowers the margin for error.
- Destination page speed: the ad is clicked in the middle of a conversation, and a slow page sends the user straight back to it. Test the mobile first screen before the ad goes live.
- Separate measurement: tag ChatGPT traffic as its own source in GA4 with UTM parameters so it does not blend with Google Ads or organic; only then can the channel's own cost per conversion be read.
- A four-week budget: the channel is read in week four, not week one. The daily floor is 25 dollars according to the press, but the decision is made on the four-week total.
The closing thesis is the sentence we opened with: a ChatGPT ad sells a space beneath the answer, not a place inside it. The first can now be rented; the second is still earned with content. The order does not change: measure first, advertise second.