ChatGPT Ads: the complete guide to launching your first campaign (2026)

What exactly is ChatGPT Ads?

When a user asks ChatGPT something like “which CRM should a five-person sales team pick?”, the assistant answers, and a sponsored unit may then appear below the answer, in a separate box. OpenAI stresses the point: ads do not influence the answer and stay visually distinct from it.

The unit is standardised: advertiser name, favicon or logo, a headline, a short description, an image and a link to your page. The system decides when to show it based on relevance between the ongoing conversation, your context hints, your ad and your landing page content.

The fundamental difference with Google Ads: you don’t bid on a query, you target an intent expressed in a conversation. The user has often described their need in detail before the ad shows up. Havas Market’s team observed ads typically appearing after the assistant’s third or fourth reply.

Who sees the ads (and who doesn’t)

This is the first thing to understand, because it defines your real audience:

• Seen by: logged-in adult Free and Go users in open markets. Since April 2026, some logged-out users too.
• Never seen by: Plus, Pro, Business, Enterprise and Edu subscribers, and accounts identified as under 18.
• Never next to: personal health, mental health or political conversations. Political advertising is not allowed.
• In Europe: personalisation based on history requires explicit opt-in consent. Without it, ads stay contextual, triggered by the current conversation only.

Open markets as of September 2026

United States (February 2026), then Canada, Australia and New Zealand (March), the UK, Japan, South Korea (May), Brazil and Mexico (summer). On August 18, 2026, OpenAI announced 31 European markets, including France, Germany, Spain, Italy, the Netherlands, Sweden, Norway and Poland, rolling out from August 24. Self-serve Ads Manager opened there on August 31, bringing the total to 52 countries according to Search Engine Journal.

Step 1: check your business is eligible

Before building anything, read OpenAI’s ad policies. Prohibited or heavily restricted: dating and adult content, alcohol and tobacco, drugs, gambling, financial services, legal services, health services and health claims, counterfeits, individual job and housing listings, political content, sensitive topics (abortion, immigration, religion, gender identity…) and unsupported wellness claims.

If you operate in a regulated vertical, expect a rejection at review, or very limited delivery even if the ad is approved.

Step 2: create and configure your Ads Manager account

Head to ads.openai.com. Access is still gradual: OpenAI verifies the business before granting full access.

Step 3: structure the campaign

The structure mirrors Meta or Google: campaign → ad group → ad. The campaign holds the objective, budget, dates and location. The ad group holds the context hints. The ad holds the creative.

Three buying objectives are available:

Step 4: write good context hints

This is the part that exists nowhere else, and the one that makes the difference. A context hint is a natural-language sentence describing a conversation situation where your offer is the right answer. OpenAI notes they guide matching without guaranteeing delivery: the system also reads your headline, description and landing page.

Weak context hint: “invoicing software”. Strong context hint: “a freelancer asking how to send compliant invoices and automatically chase clients who pay late”.

• Describe the user’s problem, not your product.
• One ad group per broad intent (compare, learn, fix an urgent problem).
• Ask ChatGPT the questions yourself: Havas Market started by auditing the assistant’s answers on its topic (moving house) before building ad groups.
• Align the landing page: if it doesn’t address the situation described, relevance drops.

Step 5: build the ad

In the European Ads Manager tested by the agency Studio en-tête, an ad has four fields: URL, a headline of up to 50 characters, a description of up to 100 characters, and a square image. In practice OpenAI recommends much shorter copy than those limits, and an image that isn’t just your logo.

A catch flagged by early testers: automatic text personalisation is on by default. Review what the system generates, or switch it off if your sector needs tight control over messaging.

Every ad goes through review before serving.

Step 6: set budget and bids

In the US, OpenAI recommends a starting bid of $3–$5 per click. In Europe, early accounts only had the “Maximise results” strategy, no manual bids, region-level geo-targeting and no custom audiences.

On budget, two schools: Studio en-tête tested with €25 a day (€220 spent in four days), while Havas Market recommends €10,000–€20,000 for a conclusive test on a large account. For a small business, €500–€1,500 over three to four weeks is enough to learn whether conversations in your category drive clicks.

Step 7: measure

Ads Manager reports impressions, clicks, spend, CTR, average CPC and CPM, plus conversions when the pixel or Conversions API is installed. Reports export to CSV. Add UTM parameters to your URLs to find this traffic in your usual analytics tool.

OpenAI never shares conversation content with advertisers: you get aggregated numbers, not the questions people asked. That’s why a landing page that qualifies visitors on its own matters.

Mistakes to avoid

Where does Perroly fit?

Perroly doesn’t create ChatGPT Ads campaigns: those are run in OpenAI’s Ads Manager. Where Perroly helps is around the campaign: spotting in your analytics the organic topics that get reactions (your best context hints), and handling the conversations that traffic generates on Instagram and Facebook with messaging automations. Meta and Google ads management is in the works; we’ll say so clearly when it ships.

To compare ChatGPT Ads with other platforms, read our 2026 ad cost benchmarks, the first hard numbers and the story of the launch.