Google invented classic search – ChatGPT is redefining it

Advertising in AI language models: does it make sense? AI platforms have very different takes on this. Perplexity pulled the plug on the experiment in February 2026, arguing that sponsored answers undermine user trust in the integrity of AI responses. Anthropic even spent a small fortune on Super Bowl spots to spread its message: "Ads are coming to AI. But not to Claude." With maximum reach.

That may have annoyed Sam Altman and market leader OpenAI, but it didn't change their decision to start testing ads in ChatGPT in early February 2026. Initially only in the US. Since then, Canada, Australia, and New Zealand have been added, and in early May OpenAI announced the launch in Brazil, Japan, Mexico, and South Korea. The UK is set to be the first European market to follow shortly. But it's not just OpenAI, Google is now integrating ads as well, currently in roughly a quarter of its AI-generated search results in AI Mode, significantly more than at the start of 2025.

Marketers should therefore be taking a hard look at visibility within AI responses, at the very latest, now. 71 percent of AI users in Germany use OpenAI's ChatGPT and 50 percent use Google's Gemini, according to a recent survey by Bitkom Research. Perplexity and Anthropic, by contrast, come in at just 7 and 5 percent respectively. And on smartphones, following Apple's announcement that it will lean on Google Gemini for AI going forward, an extremely dominant, almost monopoly-like, presence of Google's AI is taking shape.

Yes, we're currently in a transitional phase. The bulk of search traffic will still come through Google and classic search in the months ahead. And solid SEO provides an important foundation for GEO (Generative Engine Optimization, i.e., optimizing content for AI). But users are adapting to AI overviews and AI-generated answers at an unusually rapid pace. That's bad news for marketers whose brands aren't mentioned there, either in the organic response or as a paid recommendation. Because if you're not part of the answer, you're going to be less relevant going forward.

ChatGPT urgently and permanently needs money, a third revenue stream alongside subscription income from individual users and businesses. Of 900 million weekly active users worldwide (up 125 percent from a year ago), only around six percent of individual users (50 million) currently pay for a subscription.

That's why advertising in ChatGPT is now being rolled out to additional markets following the US test. After the economically strongest markets in Asia and Latin America, the European rollout is reportedly planned for the second half of this year. The UK launch was officially announced in early May. So within fewer than 90 days, advertising on ChatGPT could be live in Germany too. Unfortunately, most brands in the German market don't have a single GEO-optimized page yet, meaning the AI response engines can't process their information at all, or only poorly.

Advertising on ChatGPT isn't important simply because another major platform is selling ads. It signals a structural shift: conversation becomes the interface, and the search engine becomes the answer engine. That changes many of our longstanding rules and assumptions, because the customer decision journey can shrink dramatically. Google invented classic search. ChatGPT is burying it. A single prompt can take a user through nearly the entire marketing funnel. Conversational search, according to early findings, results in roughly 40 percent fewer touchpoints.

Users can also pause their research at any point and pick up exactly where they left off hours or days later. The consequence: a declining number of traditional search queries, fewer page visits to brand and retailer sites, and fewer opportunities to win the "click" in the traditional sense. The reason overall search traffic continues to grow is that users are now validating AI recommendations after the fact, trust in AI answers isn't yet high enough that people follow them without a quick sanity check.

For now, ads are visible "only" to ChatGPT's Free and Go users, but that's well over 90 percent of the user base. In ChatGPT, ads appear below responses when there's a relevant sponsored product or service that fits the current conversation. OpenAI states that ads operate on the principle of "response independence": ads are not meant to "influence answers" and must be clearly labeled and displayed separately. OpenAI also says it keeps user conversations confidential from advertisers and does not sell user data to advertisers. ChatGPT users can opt out of personalization and delete ad-related data.

In a chat interface, users don't prompt: "What's the best price for CRM software?" The questions are far more complex: "We're a team of 12, we're growing to 40 employees, we need GDPR-compliant tools, and vendor XY is too expensive — what software should we get?" That's not a keyword. That's a bundle of context. With conversational search, we're seeing a shift from keyword matching to "intent-bundle" targeting, where you're bidding on the context of an entire conversation rather than a single search term.

With classic search ads, you can sometimes force clicks through bids and a halfway decent landing page. With AI-driven search, you also need machine-readable content that can be reliably summarized and claims that are easy for machines to verify. In other words, proper GEO.

So far, primarily US agencies and global networks have had the chance to gather hands-on experience with ChatGPT advertising. Early reports suggest that the conversion rate for traffic coming from ChatGPT is significantly higher than Google Search's, which hovers around 3 percent. According to a Similarweb analysis from early May, click-through rates (CTR) are twice as high as display ads. Since the ads are natively embedded in ChatGPT's answers, willingness to click appears higher than with traditional banners or links. On ChatGPT, you're also no longer buying rigid keywords, you're buying "intent signals." Context beats keywords.

After the first few weeks (as of early May 2026), international trade media are reporting the following developments on ChatGPT: CPMs have dropped from $60 to $25. The minimum budget required has fallen to one-fifth of the original (from $250,000 to $50,000), and cost-per-click in testing reportedly lands between $3 and $5. The CPC model test and the reduction in minimum investment are both indicators that OpenAI isn't just going after branding campaigns from large advertisers, it's making a move into performance marketing.

In early May, OpenAI announced a beta self-serve ad manager through which US advertisers and agencies can book directly. Currently, the platform supports uploading ads, setting CPC bids, managing budgets, and tracking campaign performance. When the platform opens up to European advertisers, we should expect limited inventory, higher volatility, and a degree of chaos during the "learning phase", at least if every past launch of advertising on major digital platforms is any guide.

Marketers need to understand that they'll have to address both areas within AI overviews and responses going forward:

The organic side (GEO): Brands and companies should be appearing in responses to questions relevant to them. That primarily requires machine-readable content that AI can process quickly and effectively.

The paid side (AI ads): Brands and companies can buy their way into responses to relevant questions. That requires, among other things, new "advertising" concepts designed for dialogue with users.

To prepare well for the era of answer engines and conversational search, there are four immediate actions to take:

1. Treat GEO and paid ChatGPT as one system: Just as you would SEO and ads on Google Search.
2. Build a strong GEO foundation: Create pages with high topical authority that answer questions in an "assistant style" (e.g., comparisons, alternatives, decision guides, price breakdowns).
3. Start developing ad concepts built for conversation: Think "help me choose"-style offers (interactive/guided), comparative arguments, or messages that provide evidence and proof points.
4. Think in two scenarios: Either your brand is already part of the AI answer, or the AI answer lists only your competitors.

Of course, there are still several hurdles ahead for the launch of ChatGPT ads in Europe, particularly on the regulatory side. The main issues are existing data protection rules and the EU AI Act. Privacy experts see conflicts between advertising on ChatGPT and the GDPR, especially around the right to be forgotten. Even "anonymized" data from private AI chats can often be traced back to individuals. Brand safety is also somewhat tricky: there are no URLs to block, no publisher domains to exclude, and no pre-roll videos with clear editorial context. Brands currently have almost no control over what kinds of conversations their ads appear alongside.

But the worst outcome for brands? Not being asked about at all, because they're not part of the answer.