Checked September 3, 2026
ChatGPT Ads custom audiences and bid adjustments
Decide whether the list should exclude, include, or change a bid. Then check permission, match size, region, measurement, and rollback before upload.
On this page
Direct answer
The current rules are specific and changeable.
OpenAI currently documents custom audiences built from hashed first-party identifiers, including email, phone, and GAID. A file can mix supported identifier types. Lists larger than five million records can be split across multiple uploads. Members can be added, removed, or replaced after an audience is created.
An audience under 25,000 matched users can be used only for exclusion. Inclusion and bid adjustment require at least 25,000 matched users, and OpenAI recommends 100,000. Documented bid multipliers run from 0.1x to 10x. When more than one audience matches, the highest multiplier applies.
Connect the audience decision to paid advertising work, keep account ownership clear with the agency management guide, and follow other product changes in Learn.
Current rules
What the audience can do
| Decision | Matched-size rule | Use | Main check |
|---|---|---|---|
| Exclude | Allowed below 25,000 matched users. | Suppress customers, employees, recent converters, or another justified group. | Permission, suppression logic, freshness, and accidental over-exclusion. |
| Include | At least 25,000 matched users; 100,000 recommended. | Reach an eligible first-party audience. | Campaign objective, regional eligibility, list quality, and audience size. |
| Bid adjustment | At least 25,000 matched users; 100,000 recommended. | Change the maximum bid for a matching user. | Multiplier, overlap, highest-multiplier rule, economics, and rollback. |
File and privacy
Prepare the list before upload
- Confirm the business collected and may use the data for advertising.
- Choose supported identifiers and normalize them before hashing.
- Mix supported identifier types only according to the current file rules.
- Split a list above five million records across multiple uploads.
- Remove people covered by opt-out, suppression, or legal requirements.
- Name the source system, segment rule, owner, count, and export date.
- Record the raw count, accepted records, matched count, and audience status.
- Plan how members will be added, removed, or replaced.
- Keep proof that uploaded files are deleted after processing, usually within 24 hours under the current documentation.
- Check account and regional eligibility, including EEA and Swiss limits.
List-size estimate
Estimate raw records from the expected match rate
Required raw recordsmatched-user target ÷ expected match rate
Example only100,000 ÷ 50% = 200,000 records
This is a planning estimate. Match rate is not guaranteed. Use a rate grounded in your own identifier quality and prior uploads. Do not collect extra data merely to reach a target.
Campaign control
Tie every audience change to a business rule
Suppression
Name who should not receive the ad and why. Check that excluding past buyers does not block a renewal, replenishment, or cross-sell campaign that needs them.
Inclusion
Name the offer and action that fit the list. A large audience does not prove the message, product, or campaign is relevant.
Multiplier
Connect the 0.1x to 10x setting to margin, value, and conversion evidence. Start with a reversible test and record overlap with other audiences.
Measurement
Record delivery, cost, conversion, qualified outcome, exclusion effect, and business value. Platform activity is not the final customer record.
Common questions
Answers without a sales pitch
Can one ChatGPT Ads audience file contain different identifier types?
OpenAI’s current documentation allows supported identifier types in the audience workflow. Follow the current file schema and formatting rules before upload.
What happens if the matched audience is under 25,000?
OpenAI currently says it can be used only for exclusion. Inclusion and bid adjustments require at least 25,000 matched users.
Can I change an audience after creation?
Yes. OpenAI currently documents adding, removing, or replacing members. Keep the change date, source, accepted count, and campaign effect.
How do overlapping bid adjustments work?
OpenAI currently says the highest matching audience multiplier applies. Record overlaps before launch so a user does not receive an unintended bid.
Primary sources
What the platform documentation supports
These sources support the facts on this page. They do not disclose complete ranking systems or guarantee a recommendation, citation, click, lead, or appointment.
- OpenAI Help: Set up custom audiencesCurrent UI, thresholds, edits, and audience uses.
- OpenAI Developers: Custom audiencesIdentifier, upload, hashing, size, and API rules.
- OpenAI Help: Ads in ChatGPT basicsProduct and account context for ChatGPT Ads.
- OpenAI Advertising PoliciesAdvertising content and targeting boundaries.
- OpenAI Privacy PolicyOpenAI’s general data-processing disclosures.
- OpenAI Consumer PrivacyConsumer privacy controls and explanations.
- NIST: Protecting personally identifiable informationRisk-based handling of personal information.
- FTC: Advertising and marketing guidanceUS privacy and advertising guidance.
- European Commission: Data protectionEU data-protection framework.
- EDPB: Targeting social media usersRegulator guidance on audience targeting responsibilities.
Turn the list into a controlled campaign decision.
Stan Consulting can connect audience rules, consent, bid logic, measurement, and rollback inside a paid-media project.
Request a quote