ChatGPT Ads: What our webinar revealed

Since the end of August 2026, ChatGPT displays sponsored recommendations below responses, clearly marked as ‘Sponsored’, as the example on the right shows. The ads run separately from the language model: they cannot influence, rank, or change the actual response.

ChatGPT Ads are shown to logged-in, adult users of the Free and Go plans. The paid plans Plus, Pro, Business, Enterprise, and Edu remain ad-free.

An ad consists of:

• Name of the advertiser and small logo (favicon)
• Title and description
• Image
• Landing page URL

Product ads are automatically generated by OpenAI from the product feed. The most significant innovation compared to traditional search engine advertising is that there are no keywords. Relevance is created through the conversation context, known as Context Hints.

The figures that OpenAI uses to support the advertising launch are substantial: over 1 billion weekly active users, 1 billion US dollars in annualised advertising revenue achieved in under 200 days (as of 31 August 2026) and a revenue target of 2.5 billion US dollars for advertising in 2026.

• ChatGPT Ads are already being displayed in over 40 countries.
• Since 24 August 2026, also in 31 European markets, including Germany, Austria, and Switzerland.
• According to OpenAI, tens of thousands of advertisers worldwide are already using the format.

Sources: OpenAI (as of 31 August 2026), CNBC, Forbes, MediaPost, and Adweek.

The reach of ChatGPT Ads is smaller than in traditional search (logged-in, adult users of the Free and Go plans), but the context is significantly more precise. Users can view at any time which ad they are seeing and why.

At the launch at the end of August, bookings were still handled through the OpenAI Ads Solutions Team and partner agencies. Since the beginning of September, Self-Serve has been available via ads.openai.com, including for Germany, Austria, and Switzerland.

For the European Economic Area (EEA) and Switzerland, the following applies:

• No personalisation: Only the current conversation context counts, no old chats or saved memories.
• Digital Services Act (DSA): Transparency obligations and a public advertising archive.
• GDPR: consent-based data processing, relevant for conversion tracking.

The consequence for practice: Context Hints and landing page quality determine performance, not audience data.

Without tracking, there is no optimisation. In the OpenAI Ads Manager, the setup runs in three steps:

1. Create Pixel: under Tools → Conversions, create a new Pixel ID and embed the base code in the <head> section of each page, alternatively via the Tag Manager.
2. Define Events: Standard events such as page_viewed, items_added, checkout_started or order_created. Important: The order value is passed in cents without a comma, so €129.99 becomes 12999. Otherwise, OpenAI will count the revenue as significantly lower. An event_id (e.g. the order number) prevents double counting between the pixel and the server.
3. Conversions API setup: server-side and more reliable than the pixel alone. For oCPC (optimised cost per click), exactly one active standard event is needed as a target.

Those already using Google Tag Manager create the OpenAI Ads pixel as a tag in the web container (trigger: all pages), as shown in the screenshot on the left, along with an event tag for order_created with variables from the dataLayer. The GA4 mapping is straightforward: view_item → contents_viewed, add_to_cart → items_added, begin_checkout → checkout_started, purchase → order_created. Optional, but recommended: a server-side tag for the events API in sGTM, with the identical event_id. The Consent Mode v2 ensures that the pixel tag in the EEA and the UK only fires after consent is given.

Before the first pixel, the topic must also be included in the privacy policy:

Include OpenAI Ads pixel and Conversions API as marketing services in the consent banner, add a separate section for data protection, and name OpenAI Ireland Limited as the provider. The legal basis is consent under § 25 Abs. 1 TDDDG and Art. 6 Abs. 1 lit. a GDPR.

Important: This is not legal advice. Please coordinate the specific wording with your data protection officer or your law firm.

A campaign is created in the Ads Manager in five steps:

1. Objective and Bid: Reach (billing per 1,000 impressions), Clicks (billing per click) or oCPC (optimised cost per click) in beta.
2. Budget and Platform: Set daily budget, choose distribution on iOS, Android, and Web.
3. Targeting: Include or exclude countries, regions, or your own audience lists (Custom Audiences).
4. Ad Group: Provide context hints in plain text, as shown in the screenshot on the right.
5. Ad: Define title, text, image, and landing page URL, check preview, wait for approval.

The crucial difference from traditional search engine advertising: There is no keyword list. Instead, you describe in plain text in which conversations the ad is relevant. ChatGPT matches the ongoing context with these hints. In the EU, only the current conversation counts.

From Practice:

• Prefer several small ad groups per topic rather than one large one.
• Formulate context hints like real user questions, not like keyword lists.
• Prompts and search phrases with six or more words can be conveniently extracted from the Google Search Console using a regular expression.
• Describe situations: problem, comparison, purchase intent.
• The landing page must directly answer the question from the chat.

For product listings, create a new feed under Tools → Feeds; the Ads Manager generates a dedicated SFTP access for this, as shown in the screenshot on the left. The supported formats are Parquet (preferred), JSONL, CSV and TSV, each as gzip. XML is not supported. A daily full snapshot runs via SFTP, and price and availability can also be updated several times a day through a Delta API.

Those who already maintain the product feed for Google Shopping will be familiar with the basics: item_id, title, description, url, image_url, brand, price, availability. This is supplemented by seller_name, seller_url and return_policy as well as three yes/no switches per product:

• Is it allowed to be advertised? (is_ads_eligible)
• Is ChatGPT allowed to show it in responses? (is_eligible_search): if enabled, the resulting clicks are free.
• Can it be purchased directly in the chat? (is_eligible_checkout)

As the feed is Google-compatible but not identical, an automated intermediate step is worthwhile. Our Rheinwunder Feed Manager reads the existing Google Shopping feed, maps it to the OpenAI schema, checks a safeguard against erroneous runs, and automatically transfers the result daily at 4 am via SFTP as CSV.GZ. The result: no duplicate feed maintenance and product listings from day one.

Generative Engine Optimisation (GEO) ensures that content from ChatGPT, Gemini, Claude, and Google AI Overviews is found, cited, and recommended. The difference from ads can be summarised simply: Ads buy presence in chat, GEO earns it. Together, this creates a dual presence in the same conversation, and brands that are already familiar with AI systems appear more trustworthy in ads as well.

Without the technical and content SEO fundamentals, there is simply nothing for an AI to cite. AI responses rely on crawled, indexed, and comprehensible sources:

• Crawlability: GPTBot, OAI-SearchBot, and others must not be blocked.
• Structure: clear H1/H2, short paragraphs, answer first, takeaways at the end.
• Structured Data: Schema.org markup for Organization, Article, FAQPage, and Product.
• E-E-A-T: visible authors, date, sources, and author profiles.

And it is precisely this quality that benefits doubly: The landing pages of your ChatGPT ads are supported by the same SEO foundation. How prominently a brand is currently visible in AI responses can be continuously measured with our tool appariq, including Share of Voice, sentiment, and competitive benchmark, as shown in the overview on the right.

1. ChatGPT Ads are here: available in DACH since 24 August 2026, and available for self-booking since early September.
2. No tracking, no optimisation: set up Pixel and Conversions API before the first campaign.
3. Context hints instead of keywords: think in conversational situations, not in search terms.
4. Product feeds are Google-compatible, but not identical: an automated feed manager saves manual work and errors.
5. GEO builds on SEO: those who are cited organically win double in chat.

Do you not wish to implement these steps yourself, but prefer to have campaigns, tracking, and feeds managed from a single source? On our performance page for ChatGPT Ads we show how collaboration with Rheinwunder looks.