Get started
Browse guides
Guides

OpenAI Ads and operating rules

Create copy from context and product evidence, then inspect ad performance.

4 min readDocument reviewed

Before you start

You need Pro or Enterprise and a compatible OpenAI Ads account. Check the product feed and permissions. Hosted jobs require AI credits.

Step by step

  1. Check the account and product feed under Ad settings. For review-based ads, also inspect product associations in Reviews.
  2. Select reviews or a brief, the product, and the ad skill. Begin with a small scope to inspect output quality.
  3. Read the copy, product evidence, and evaluation results. Correct exaggerated wording or unsupported efficacy claims.
  4. Check the ad set, campaign, and output settings. Supported workflows may allow paused, active, export, or hold outputs; verify destination and state before running.
  5. Set supported spend, impression, click, CPC, or CTR conditions and minimum samples under Automatic rules. Preview the rule and check ad, group, or campaign scope before enabling it.
  6. Review performance and rule history together. Marketing pause rules are separate from the three supported hosted Automations jobs.
  7. Use Ad Performance to compare context-linked prompts and ads for the same product, market, and period. This report requires Pro or above. Treat uncollected values as missing, not zero.

Check the result

Confirm the evidence behind the ad and the state in which it was delivered. Distinguish impressions, clicks, and spend from actual conversion measurement.

Context performance

Open Ad Performance. Check Pro-and-above access, the connected ad account, and actual collection scope.

Connect one buying context from a review to a tracking prompt and an ad, then compare organic mentions and paid results for the same product, market, and date range.

  • Context performance: inspect observed brand mention rate, the answer count, and average rank among answers mentioning the brand alongside CTR, CPC, and ad impressions.
  • Organic and paid: compare results by context to plan content and ad tests. Observational comparisons do not establish advertising’s causal effect.
  • Tracking gaps: distinguish prompts without ads, ads without prompts, and results not yet collected. Unmeasured values are not reported as zero.

Brand mention rate uses collected AI answers as its denominator. It is not a recommendation probability across all AI users or a probability of an ad being served.

If something goes wrong

  • Check connection state and error details for account, product approval, or API access problems.
  • If a rule does not run, check whether it is enabled, its scope, minimum sample, and history.
  • Do not fill unavailable ROAS metrics with estimates.

Related guides