ChatGPT Ads Are Live in Europe. Here's What Actually Changes for Your SEO.

On August 24, 2026, ChatGPT started showing ads to people in 31 European countries. Germany included. Within about a week I had three separate versions of the same client question in my inbox, and they all boiled down to: do we need to get in before the prices go up?

Fair question. Wrong panic.

I have spent the two weeks since then reading OpenAI's own documentation, the EU's designation notice, and roughly a dozen agency blog posts claiming to have the real cost-per-click numbers. Here is what I actually found, and what it means for the SEO work you are paying someone to do.

Short version: ChatGPT ads are real, the ads manager is real, and the SEO impact is close to zero this quarter.

The ad unit sits below the answer, not inside it

This is the single most important design detail and most of the coverage skips past it.

ChatGPT ads appear underneath a completed response, as a visually separated card labelled "Sponsored." They are not injected into the paragraph. They are not a brand name swapped into a recommendation. They are not a link inside the answer text that happens to be paid for. The model writes its answer, and then a separate system attaches a commercial unit below it.

Compare that to a Google results page, where the paid block sits above the organic results and pushes them below the fold on mobile. In ChatGPT the paid unit sits after the thing the user actually came for. Whether that placement holds forever is a different question, and I will get to it.

Ads only show to free tier and Go plan users. If you pay for Plus or a business seat, you see none of this. That matters when you picture your audience, because the buyer who asks ChatGPT to compare enterprise CRM vendors is more likely to be on a paid seat than the person asking which running shoes to buy.

The targeting signals reported so far are also narrower than most people assume. At launch in Europe, ads are matched on what you are currently discussing plus approximate location, device type, time of day, and language. Personalization from past chats requires an explicit opt-in under GDPR, and advertisers get no access to your chats, chat history, or memories. Advertisers can supply what OpenAI calls context hints, essentially a text box describing the conversations and topics where they think their product fits.

So it is closer to contextual placement against a whole conversation than to a keyword auction. Sensitive categories are excluded: no ads near personal health, mental health, or political topics, and political advertising is not permitted at all right now.

OpenAI says ads do not influence answers. You cannot verify that.

OpenAI's position is unambiguous. Ads run on separate systems from the chat model. Advertisers cannot shape, rank, or alter ChatGPT's responses. The sponsored card is labelled and visually separated from the organic answer.

I believe that is true today. I also want to be honest that neither you nor I can check it.

With Google, you can. Ads and organic are separately observable on the same page, the same query run from a clean browser gives a comparable result, and thousands of people rank-track the same terms daily. If Google quietly started favoring advertisers in organic, someone would find the pattern in the data.

ChatGPT gives you none of that. Answers are non-deterministic. The same prompt produces different wording and often different citations on two consecutive runs. There is no shared index everyone can query. There is no rank tracker measuring position one. If OpenAI ever softened the wall between the ad system and the recommendation logic, the honest answer is that the SEO industry would probably notice late, and would notice through anecdote rather than proof.

That is not an accusation. It is a structural fact about the medium, and it is the reason I keep telling clients that AI visibility work is a slower, noisier feedback loop than search. I wrote about that measurement problem in more detail in GEO Is the New Snake Oil (Or Is It?), and nothing about the ad launch changes the underlying issue.

Europe got the ads, then Brussels got ChatGPT

The regulatory timing here is worth understanding, because it is the part that might actually constrain what OpenAI does next.

The UK went first, on June 2, 2026, outside the GDPR perimeter. The 31 EU and wider European markets followed on August 24. Then on August 31, the European Commission designated ChatGPT a Very Large Online Search Engine under the Digital Services Act, citing 159.1 million monthly active users in the EU against a 45 million threshold.

That designation carries obligations that land on the ad business directly. Very large services must maintain a public advertising repository showing each ad's content, its sponsor, how long it ran, the principal targeting criteria, and how many users it reached. They face an annual independent audit and must provide data to regulators and vetted researchers. Non-compliance tops out at fines of six percent of global annual turnover. OpenAI has until the end of November 2026 to comply, supervised by the Commission alongside Ireland's Coimisiún na Meán.

Here is why that is genuinely interesting for us rather than just legal trivia. An ad repository is a public dataset. If it ships as specified, you will be able to see which competitors are buying, roughly what topics they are targeting, and how long they sustained a campaign. That is a competitive intelligence source that does not exist for Google Ads at anything like that granularity. Set a reminder for December.

On the consent side, OpenAI leans on legitimate interest for generic contextual ads and explicit consent for personalized ones. Practically, that means European users can change which ads they see but cannot turn ads off without paying. Expect at least one privacy complaint about that framing. I would not build a strategy around it disappearing.

There is an ads manager, and you can probably get in

Two weeks ago the honest answer to "can I buy ChatGPT ads" in Europe was "only through OpenAI's sales team or an agency partner." That changed on August 31, when a beta self-serve Ads Manager opened to eligible advertisers across the 31 European markets at ads.openai.com.

Bidding covers traffic and conversion objectives. There are custom audiences, a pixel, and a conversions API. The managed routes stayed open too, with OpenAI's own sales team plus agency holding companies and adtech partners still booking directly.

So access is no longer the blocker. Which brings us to what those clients were really asking.

Nobody has trustworthy cost data yet, including me

I have never run a ChatGPT ad campaign. Not one euro. I am not going to pretend otherwise so this post sounds more authoritative.

What I did instead was collect every published benchmark I could find. The results are a mess. One agency reports roughly $1.72 average CPC from fifteen days of live spend. Another publishes $3 to $5 for ecommerce and $8 to $18 for software and finance. Click-through rate estimates float between 0.5 and 1.8 percent. Conversion rate claims range from 4 to 7 percent for commercial verticals.

Those cost figures disagree by nearly an order of magnitude. Every one of them comes from a company that sells ChatGPT ad management. Most cover two to six weeks of spend in a brand new auction where demand has not caught up with inventory, which mechanically suppresses prices in a way that will not last.

The most useful sentence I can write about ChatGPT ad economics in September 2026 is this: nobody has enough data yet to tell you what a click is worth, and anyone quoting you a confident CPC benchmark is quoting a number with a two-week sample behind it.

For context, and I am deliberately keeping this to one paragraph because it is a whole other post: OpenAI's ad business hit a $1 billion annualized run rate on August 31, less than 200 days after launch, against a reported internal target of $2.5 billion for the year. Fast growth, still behind plan. That tells you the pressure to expand ad surface area is real. It does not tell you what your click costs.

Does any of this change the SEO work? Mostly no.

I keep looking for the part where ChatGPT ads change the SEO work, and I keep not finding it.

Your organic visibility in ChatGPT still comes from the same place it came from in July. The model surfaces and cites sources it can retrieve, parse, and trust. Nothing in the ad system touches that pathway. Adding a sponsored card below an answer does not make your unlinked brand mentions matter less, does not change how your pages get chunked and retrieved, and does not create a paid shortcut into the answer itself. The mechanics I walked through in How to Get Cited by ChatGPT are identical this week to last week.

The one honest second-order effect is attention. A sponsored card below the answer competes for the click that used to go to a citation, on the subset of conversations where a relevant advertiser exists. That is real but small, and it lands on free tier traffic specifically. If you sell to businesses, a meaningful chunk of your ChatGPT audience is on a paid seat and sees nothing.

The other thing worth noticing is a quiet alignment. The stuff that gets you cited organically, clear pages that answer one question properly, specific language that matches how people actually phrase a problem, is the same stuff that makes a paid landing page convert against a conversational prompt. If you were going to test ads eventually, the prep work is the content work you should be doing anyway. That is convenient, not urgent.

What I told all three of those clients was the same thing: keep the retainer pointed at the citation and coverage work, and revisit ads in Q1, once there is a benchmark set worth trusting and the DSA repository is live. Nobody argued. This is not a hard call.

The reason I am confident is the same reason I was skeptical of the last three AI search panics. Google published its own guidance on ranking in AI experiences and it read like standard technical SEO with a fresh coat of paint, which I picked apart in Google Just Told You How to Rank in AI Search. And when Search Console finally started reporting AI traffic, my own numbers were far less dramatic than the discourse suggested, which I laid out in Google Search Console Now Reports AI Traffic. The pattern repeats. A visible change to the interface gets read as a change to the underlying mechanics, and usually it is not one.

What I am watching, and what would change my mind

Four things, in the order I expect them to matter.

The DSA ad repository, due by end of November. Whether it actually ships with per-ad targeting criteria and reach data, and whether it is queryable or a PDF dump. If it is good, it becomes a competitive research tool nobody else has learned to use yet.

Ad placement creep. Right now the sponsored card sits below the answer. The moment OpenAI tests a unit above the response, inside a product comparison, or blended into a shopping flow, my "mostly no" becomes "we need to talk." Screenshot placements periodically. Placement drift is the leading indicator here, not pricing.

Ads reaching paid tiers. Ads are on free and Go today. If they appear for Plus users, the reachable audience changes shape and B2B calculations change with it.

A CPC dataset with more than one quarter behind it. Not an agency blog post. Something with enough accounts and enough months to survive the auction maturing.

And the thing that would genuinely change my mind: any credible, reproducible evidence that advertiser status correlates with organic citation frequency. Not a vibe, not one screenshot, a study across enough prompts to rule out noise. If that ever lands, the entire recommendation in this post is wrong and I will say so in a follow-up. Until then, the reasonable move is to keep the wall in your head where OpenAI says it is in their systems, and keep doing the work that earns citations.

One more thing worth doing while you wait. Run twenty prompts a customer would plausibly type, in a logged-out session on the free tier, and record what comes back: whether you are cited, who else is, and whether a sponsored card appears at all. That takes an afternoon and gives you a baseline. In six months it will be the only before-and-after data you own. If your brand is showing up in those answers described in ways you did not choose, that is a separate and more urgent problem, and I covered how to handle it in Your Clients Are Googling Your Brand and Getting an AI Summary Instead.

The channel is new. The work is not. If you want a second opinion on whether your site is actually retrievable and citable before you spend anything on a new ad platform, that is most of what I do at Wunderlandmedia.