ChatGPT Ads: How to set up your first test and decide whether to continue?

Since August 24, 2026, French advertisers have been able to run ads on ChatGPT, and as of August 31, any business can open a self-service account. Many marketing teams are asking the same question: should we test it now, what should the budget be, and how do we know after a few weeks if it's worth continuing?

At Growth Room, we have been supporting advertisers on ChatGPT Ads since the platform launched in France. Here is the method we use to build a clean initial test, for both B2B and e-commerce, and the criteria we use to make a final decision.

Key takeaways

• ChatGPT Ads has been open in France since August 2026 as a self-service platform, featuring targeting based on conversation context.
• A serious test requires a media budget of approximately €2,500 per month for 6 to 8 weeks.
• The OpenAI pixel and Conversions API must be set up before the first impression.
• The test is evaluated based on cost per qualified lead or ROAS, compared to Google Ads and Meta.
• At the end, there are three possible outcomes: continue, adjust and relaunch, or stop.

Why test ChatGPT Ads now?

There are three reasons to launch a test early.

• Competition is still low. Few advertisers are present, and cost-per-click remains low for now.
• The expressed intent is rich. A user asking ChatGPT "which payroll software should I choose for a 50-employee SME" reveals much more than three keywords typed into Google.
• The platform is growing fast. In late August 2026, OpenAI announced it had surpassed $1 billion in annualized advertising revenue in less than 200 days. Advertisers who learn to use it early will have a head start.

Two limitations to keep in mind: ads are only shown to users on Free and Go plans (not to Plus, Pro, or Enterprise subscribers), and the format is short, with a headline of about 16 characters and a description of about 32 characters. For a detailed breakdown of how the platform works, we have published a complete guide to ChatGPT Ads.

What should you prepare before launching a ChatGPT Ads test?

Half of all failed tests happen before launch. Before spending your first dollar, check these points:

• an account on the OpenAI Ads Manager (ads.openai.com);
• the pixel OpenAI pixel installed on your site and connected to the events that matter (lead, demo request, purchase);
• the Conversions API connected, so you don't rely solely on cookies;
• conversion tracking set up in your CRM (like HubSpot) and in GA4;
• a landing page that aligns with the original conversation, not just your homepage;
• a benchmark: your current cost per lead or ROAS on Google Ads and Meta.

Without this last point, you will never know if ChatGPT Ads is "expensive" or "profitable."

How do you set up your first ChatGPT Ads test?

What goal should you set for a ChatGPT Ads test?

A single primary goal, chosen before launch:

• in B2B: the cost per qualified lead (MQL or meeting), measured in the CRM;
• in e-commerce: the ROAS or customer acquisition cost.

Impressions, clicks, and click-through rates are used to manage the test on a daily basis. They are not used to decide whether to continue.

Which conversation contexts should you target?

ChatGPT Ads does not work with keywords. Conversational targeting is based on context hints : you describe the conversational situations in which your ad makes sense. Since July 2026, you can also import audience lists to include or exclude contacts, such as your current customers.

A few examples of contexts:

• in B2B SaaS: "comparing expense management software for an SME," "automating follow-ups for unpaid invoices";
• in e-commerce: "gift ideas for a cooking enthusiast," "how to choose a mattress for back pain."

For a first test, start with 3 or 4 distinct contexts, drawn from the questions your sales or customer service teams hear every week.

How should you structure the campaign?

We generally build 2 to 4 ad groups, one per conversation context, with 2 or 3 ads in each. A simple rule: test only one variable at a time. If you change the context, the message, and the landing page all at once, you won't know what made the difference.

The campaign uses the "Conversions" objective so that the algorithm optimizes bids based on your conversion events rather than just clicks.

How do you write such short ads?

With about 16 characters for the headline and 32 for the description, every word counts. What works best in this format:

• use the vocabulary from the user's question;
• focus on a concrete benefit rather than a slogan;
• include a quantified proof point when space allows (e.g., "30-day trial," "20% off your first month");
• prioritize high-quality imagery, as it does much of the heavy lifting in sponsored cards.

What budget and duration should you plan for?

OpenAI requires a minimum of €25 per day, or about €750 per month. This is enough to gain visibility, but not enough to draw meaningful conclusions. We recommend a media budget of at least €2,500 per month for a period of 6 to 8 weeks. This gives the algorithm time to learn and then generate enough conversions to compare ad groups against one another.

Which metrics should you track during a ChatGPT Ads test?

During the test, we monitor two categories of metrics. Operational metrics (impressions, click-through rate, cost per click, landing page conversion rate) are used to refine ads and contexts on a weekly basis. Decision-making metrics, however, are only evaluated at the end.

How do you decide whether to continue with ChatGPT Ads after the test?

At the end of the 6 to 8 weeks, we make a go / no-go decision based on one primary criterion: the cost per useful conversion compared to your other channels.

• Continue and increase the budget if the cost per qualified lead or ROAS is equal to or better than your performance on Google Ads and Meta, or if it is approaching those levels with a clear upward trend.
• Adjust and launch another cycle if one or two ad groups stand out while others are dragging down the average. We cut the underperforming contexts and double down on the successful ones.
• Stop or pause if no context produces actionable conversions, despite having solid ads and a well-optimized landing page. This happens when your offer is rarely mentioned in conversations with ChatGPT.

A piece of advice from our experience: don't judge a channel based on its first two weeks. When we restructured Poiscaille's Google Ads and Meta campaigns, the cost per subscription on Google dropped from €77 in November to €6 in January, and from €112 to €27 on Meta, thanks to continuous testing of audiences and creatives. In total, Poiscaille's acquisition cost was divided by 7 in 3 months, with over 2,000 subscriptions generated. We apply that same patience and rigorous testing methodology to ChatGPT Ads.

In B2B, decisions should be driven by CRM data. That’s exactly what we did for Airties: by optimizing for MQLs rather than lead volume, the cost per MQL dropped to nine times below the target by the third month.

What mistakes should you avoid during your first ChatGPT Ads test?

• Launching without a pixel or Conversions API, then judging the test based on clicks.
• Sending traffic to the homepage instead of a page that answers the specific question asked.
• Testing ten contexts with €750 per month: none of them will have enough data.
• Changing ads every two days during the learning phase.
• Comparing ChatGPT Ads to Google Ads based on cost-per-click rather than cost-per-customer.
• Forgetting to exclude your existing customers using audience lists.

Before spending a single euro, write down your decision criteria

A successful ChatGPT Ads test is largely won before launch: a single objective, conversation contexts chosen based on your customers' actual questions, tracking linked to your CRM, and a decision threshold written in black and white. With these foundations, six to eight weeks are enough to know if the platform deserves a permanent place in your mix. Our team ChatGPT Ads handles this scoping, technical setup, and test management, right up to the final decision.

‍