Where the opportunity starts
Someone researching a decision can encounter a relevant advertiser while using ChatGPT. That is a potential moment to explain a useful offer, rather than an entitlement to appear beside every conversation about the category.
OpenAI currently describes eligible Free and Go inventory, with higher paid and business plans excluded. Availability depends on rollout and account conditions. Check the actual market before planning a campaign around an assumed audience size. An advertiser's ability to open an account does not establish every country or user it can reach.
Sources: OpenAI: Ads in ChatGPT · OpenAI: Advertiser basics
The campaign has three connected levels
A campaign contains the commercial goal and spending controls. Its ad groups organize related needs or themes. Individual ads carry the message, image and destination. OpenAI's quickstart describes this structure alongside account verification, billing and ad review.
For an illustrative software business, a general product campaign might contain separate groups for a first-time setup and a migration. The distinction is useful when those buyers need different explanations. Creating separate groups for every wording variation would add administration without necessarily creating a clearer test.
Sources: OpenAI: First campaign quickstart
Context hints add meaning
A hint explains an accurate feature, customer need or situation at ad-group level. OpenAI's framework asks what the offering is, who it helps and when it is useful. Each phrase can focus on one of those dimensions. Hints are relevance inputs rather than exact-match keywords or delivery instructions.
An illustrative hint could explain that a tool helps a small operations team replace manual weekly reporting. The landing page would still need to show what the tool actually does. Repeating a category name in increasingly long lists gives a buyer less useful information than explaining one concrete need.
Sources: OpenAI: Context hints
The destination finishes the message
OpenAI recommends choosing the destination closest to the advertised offer. A relevant page should help the person evaluate what happens after the click. Its promise, price or availability, where stated, should remain consistent with the ad.
For a service, a useful page can explain the work included, the decisions the client needs to make and the next action. Show important meaning as readable page content. The ad-review crawler also needs access; a page hidden behind authentication or an automated challenge can interrupt review.
Sources: OpenAI: Creating ads · OpenAI: Crawler access
Design the test around a business decision
Our recommended planning sequence is to define a buyer situation, choose a matching destination, identify the real conversion and set a bounded experiment. Decide in advance what would justify increasing the test, changing its message or stopping it. This is a decision framework, not a claimed platform performance result.
For lead generation, keep a visit, an accepted contact enquiry and a confirmed meeting separate. Review the enquiries themselves when permission and systems allow. A large click count does not answer whether the service attracted suitable prospects, and a tracking tag alone does not establish qualified revenue.
Keep a record of the settings and creative in force during the test. Changing the destination, offer and measurement simultaneously makes a result harder to interpret. Learn from a small number of meaningful angles before increasing complexity.
What remains uncertain
The platform is evolving, and OpenAI does not publish universal performance benchmarks across industries. Treat claims about guaranteed volume or cheaper acquisition with care. For help reviewing a specific setup, see Crispy Scale's ChatGPT Ads service at ChatGPT Ads management. The starting point is the current offer and account, followed by a clear measurement plan.
Sources: OpenAI: Advertiser FAQ
How this guide was prepared
Platform mechanics are based on the official sources below. Test-design recommendations are Crispy Scale’s evaluation method. Illustrations are examples; they are not client campaign results.
Sources checked
Official documentation checked . Platform details can change.