Marketing Pulse | AI advertising
ChatGPT Ads now accepts customer lists. Who should you include, exclude, or bid higher for?
OpenAI now documents custom audiences built from customer or prospect lists, campaign inclusion and exclusion, and ad-group bid multipliers from 0.1x to 10x. The controls are specific. The audience decision still belongs to the advertiser.
Stan Consulting verdict
A matchable list is not automatically a valuable audience.
What changed: OpenAI now documents custom audiences created from email addresses or phone numbers. An advertiser can use a ready audience to limit a campaign, exclude a group, or change an ad group’s maximum bid when a matched person sees the ad.
The controls create three different marketing decisions. Inclusion says the campaign should serve only to the selected group. Exclusion says the selected group should not see the campaign. A bid multiplier says a matched person may be worth more or less in this ad group.
Those decisions require more than a clean file. The advertiser needs the person’s relationship to the business, lifecycle stage, campaign purpose, permitted use, suppression state, and the business outcome that will judge the spend.
Do not use a bid multiplier to avoid deciding what the audience is for.
A list can contain enough matched users and still combine current customers, prospects, recent purchasers, inactive contacts, and people who should be suppressed. Size makes the audience usable. It does not make the campaign coherent.
Three controls
Choose the audience action before the upload.
Include
Campaign eligibility is limited to people in one or more selected ready audiences.
- Use when the campaign has a defined known group.
- State the offer and lifecycle stage that make the group relevant.
- Confirm the final audience remains above the matched-user floor.
- Do not mix unlike groups only to make the list larger.
Exclude
Campaign eligibility remains broad except for people in the selected audience.
- Use to protect recent purchasers from acquisition offers.
- Separate employees, partners, ineligible contacts, or suppression groups.
- Keep the exclusion current when the customer record changes.
- Confirm the remaining eligible audience still meets the minimum.
OpenAI documents multipliers from 0.1x to 10x at the ad-group level. Before raising or lowering a bid, write why this matched group should have a different maximum bid, which outcome will prove it, and the condition that will reverse the change.
Current platform mechanics
The audience must clear the platform floor and the business test.
| Current rule | OpenAI documents | Advertiser decision |
|---|---|---|
| Minimum ready audience | At least 25,000 matched users | Do not dilute a useful segment with unrelated contacts merely to clear the floor. |
| Recommended size | At least 100,000 users | Treat this as delivery guidance, not proof that a larger blended audience is better. |
| Upload capacity | Up to 5,000,000 identifiers in a file up to 500 MB | Use one identifier type per upload and remove duplicates and invalid records before interpreting match size. |
| Audience changes | Custom audiences cannot be edited after creation | Create a dated replacement, archive the old audience, and record which campaigns used each version. |
| Processing | Processing commonly takes 20 to 30 minutes and can return Ready, Too small, or Failed | Save the processed result. Uploaded rows and matched users are different counts. |
Audience plus conversation
A customer list and a context hint answer different questions.
OpenAI says context hints describe the conversations, topics, needs, or situations where an ad may be relevant. They guide matching but are not exact-match targeting rules.
A custom audience describes a known relationship to the advertiser. A context hint describes the conversation where the offer may fit. One person can belong to the audience and still be in the wrong conversation for the ad. One conversation can fit the offer while the advertiser still needs to exclude a recent purchaser.
Customer-list question
Who is this person to the business, and should that relationship change eligibility or the bid?
Context-hint question
Which needs, situations, and themes make this ad relevant inside a conversation?
OpenAI recommends separate ad groups when products, audiences, or use cases are meaningfully different. That supports a simple rule: if the audience needs a different reason, offer, page, or success measure, it probably needs a separate ad group.
Use this now
The 15-minute customer-list audience check.
This saves progress in this browser. It does not send list or account information to Stan Consulting.
No audience check recorded in this browser yet.
Decision guide
Match the customer condition to the campaign action.
| Customer condition | Possible action | Proof to require |
|---|---|---|
| Recent purchaser seeing a first-purchase offer | Exclude from the acquisition campaign | Lower existing-customer overlap without losing valid new-customer delivery. |
| Qualified open opportunities with one relevant offer | Test inclusion in a separate campaign | Opportunity movement, qualified conversion, and sales outcome beyond ad delivery. |
| High-value repeat buyers for a relevant new offer | Test a separate ad group and a written bid hypothesis | Incremental repeat revenue or margin compared with the declared baseline. |
| One large file mixing customers, prospects, and unknown contacts | Do not upload yet | Lifecycle groups, ownership, suppression, and campaign purpose become known first. |
| Useful segment cannot reach 25,000 matched users | Use another test design or wait | No unrelated contacts were added simply to clear the platform floor. |
Ownership boundary
The upload needs a business owner before it needs a bid.
The operator preparing the file should know where the identifiers came from, who owns the record, what permitted use applies, which contacts must be suppressed, and who will answer a customer or internal question about the audience.
OpenAI documents plain and SHA-256 hashed identifier formats. File formatting is only one part of responsible use. Hashing does not decide whether the business should upload the data, include the person, exclude the person, or use the match to change a bid.
Keep the decision record small and complete: source, owner, date, identifier type, segment definition, suppression rule, audience action, campaign or ad group, multiplier, outcome, review window, and archive owner.
If permitted use, consent, suppression, category restrictions, or local requirements are unclear, pause the upload and send the question to the appropriate privacy, legal, compliance, or data owner.
Original sources
Current OpenAI documentation.
- Set up Custom Audiences for your Campaign for identifiers, file limits, processing, matched-size thresholds, inclusion, exclusion, bid multipliers, and replacement rules.
- Create Ad Groups for ChatGPT for context hints, ad-group separation, and the distinction between broad thematic guidance and exact-match targeting.
- Ads in ChatGPT: The Basics for current ad delivery, campaign context, and landing-page inputs.
- OpenAI Ad Policies for current advertiser, placement, category, and safety requirements.
Connected decisions
Audience targeting needs measurement and a clean customer record.
Need the audience, campaign, and customer outcome connected?
Build the paid-media decision before the upload.
Stan Consulting connects audience planning, paid ads, CRM records, landing pages, conversion tracking, and customer quality.