Audience creation in OpenAI Ads should be approached somewhat differently from the interest-based or demographic targeting models used by traditional advertising platforms.
Within ChatGPT, users often describe a need, compare products, or ask for recommendations before making a decision. For this reason, two areas should be considered together when developing a targeting strategy:
- The brand’s own customer or prospect lists
- Context hints that describe the user’s conversational context
Custom Audiences allow you to use existing customer or prospect lists in ChatGPT Ads campaigns. Context hints explain the types of conversations in which an Ad Group may be most relevant.
What is a Custom Audience?
A Custom Audience is an audience created using a brand’s own customer or prospect lists.
In OpenAI Ads, email addresses or phone numbers can be used to create Custom Audiences. These lists can be applied at the campaign and Ad Group levels for different purposes.
Custom Audiences can be particularly useful in the following scenarios:
- Showing exclusive campaigns to existing customers
- Targeting users who previously submitted a lead form
- Using more competitive bids for high-value customer segments
- Excluding users who have already completed a purchase
- Reducing bid levels for lower-priority audiences
This structure allows OpenAI Ads campaigns to move beyond broad targeting and become more controlled through the use of the brand’s own first-party data.
Strategic recommendation: Before uploading a Custom Audience, do not treat your customer database as one large file. Creating separate segments based on customer value, purchase date, category interest, or lead quality can make campaign management more flexible.
How can Custom Audiences be used?
As of July 2026, Custom Audiences in OpenAI Ads can be used for three main purposes.
1. Campaign-level targeting
The Include audience option allows a campaign to target only users included in a specific Custom Audience.
This strategy can be used for:
- Existing customer campaigns
- Loyalty program members
- Users who previously submitted a lead form
- Users who showed interest in a specific product category
- High-potential customer lists
For example, you can create a dedicated campaign for users who previously submitted a quotation form on your website.
2. Campaign-level exclusion
The Exclude audience option allows specific users to be excluded from a campaign.
This strategy can be useful for:
- Excluding users who purchased within the last X days
- Preventing existing customers from seeing acquisition campaigns designed for new customers
- Excluding users who are not eligible for a specific campaign
For example, in a new customer acquisition campaign, existing customers can be excluded so that the budget is directed toward a more relevant audience.
3. Ad Group-level bid multiplier
Custom Audiences can also be used with bid multipliers at the Ad Group level. This setting helps increase or decrease the maximum Ad Group bid for users who match the selected audience.
For example:
- You can double the bid for high-value customers by using a 2x bid multiplier.
- You can increase the bid fivefold for loyal customers identified through RFM segmentation by using a 5x bid multiplier.
- You can reduce the bid by half for lower-priority audiences by using a 0.5x bid multiplier.
The important point is that a bid multiplier does not independently determine who will see the campaign. It only increases or decreases the bid for users who match the selected Custom Audience.
Strategic recommendation: Do not automatically assign high bids to every audience. Use higher bid multipliers only for segments that are genuinely valuable, have a strong likelihood of conversion, or represent a strategic priority.
How should the list be prepared before uploading a Custom Audience?
As of July 2026, according to OpenAI’s documentation, Custom Audiences can be created using CSV or TXT files. A file can be up to 500 MB and contain up to 5,000,000 identifiers in a single upload.
The main rules to consider when preparing the list are:
- Only one identifier type—email address or phone number—should be used in each upload.
- The identifier type can be email, phone number, SHA-256-hashed email, or SHA-256-hashed phone number.
- TXT files should contain one identifier per line.
- CSV files can be uploaded with or without a header. If a header is used, the column name must be written correctly.
- Accepted column names are:
email,phone_number,email_sha256, andphone_number_sha256. - A Custom Audience cannot be edited after it is created. If the list needs to change, a new audience should be created and the old audience should be archived.
It is therefore important to clean the list and prepare it in the correct format before uploading it.
What should be considered when formatting emails and phone numbers?
Incorrect formatting is one of the most common problems during audience uploads.
For email addresses:
- A valid email address should be used.
- The email address should contain only one
@symbol. - Uppercase letters are accepted and normalized to lowercase.
- Leading and trailing spaces are removed.
- The email address should not contain spaces.
For phone numbers:
- E.164 format should be used.
- The number must include
+and the country code, such as+905321234567. - Phone numbers without a country code are not accepted.
- Characters such as spaces, hyphens, parentheses, and periods may be normalized.
If hashed data is used, the SHA-256 format must be followed. The hashed value should be a 64-character SHA-256 hexadecimal digest.
Strategic recommendation: Clean the list in your CRM or CDP before uploading it. Invalid emails, missing country codes, duplicate records, and outdated customer data can reduce the match rate. As a result, you may not be able to target every user in a list containing 50,000 records.
How do you create a Custom Audience in OpenAI Ads?
The process of creating a Custom Audience in OpenAI Ads Manager generally consists of the following steps:
- Go to the Tools section in Ads Manager.
- Open the Audiences tab.
- Select Create custom audience.
- Enter a descriptive name for the audience.
- Select the identifier type: Email, Phone, Hashed Email, or Hashed Phone.
- Select and upload your audience file.
- Create the audience by selecting Create audience.
After the audience is created, the file is processed. Processing time may vary depending on file size. OpenAI’s documentation states that this process may generally take 20–30 minutes.
What does the audience status mean?
After creating a Custom Audience, the audience status should be monitored in Ads Manager.
The main statuses are:
- Processing: The file is being processed and the audience is not yet ready to use.
- Ready: The audience has been processed, meets the minimum matched-user requirement, and can be used.
- Too small: The audience does not meet the minimum requirement of 25,000 matched users and cannot be used.
- Failed: The file could not be processed. The format and content should be reviewed before uploading it again.
The number of uploaded rows may not be the same as the number of matched users. Invalid values, duplicate records, or unmatched identifiers are not included in the final audience size.
Why does minimum audience size matter?
A Custom Audience in OpenAI Ads must contain at least 25,000 matched users before it can be used. OpenAI recommends audiences containing at least 100,000 users for more effective use.
This means that very small CRM lists may not be eligible for use as Custom Audiences.
For example, if you upload an email list containing 30,000 rows, the final matched audience may fall below 25,000. In this case, the audience will receive a “Too small” status and cannot be used in campaigns.
Strategic recommendation: Do not evaluate list size only according to the number of raw records. After invalid, duplicate, and unmatched entries are excluded, the usable audience size will generally be smaller.
What are Context Hints in OpenAI Ads?
Context hints are signals that describe the conversational situations in which an Ad Group may be relevant. They should not be treated as exact-match keyword targeting.
The purpose of Context Hints is to describe to the OpenAI advertising system the user need, problem, or decision-making process in which encountering your brand may be relevant.
For example, instead of entering only “running shoes” as a context hint, the following statement would be more descriptive:
“Users who are new to running and are looking for comfortable, supportive running shoe recommendations for daily training.”
This statement explains more clearly what the product is, who it is suitable for, and in which use case it becomes relevant.
The difference between Context Hints and Custom Audiences
Custom Audiences allow brands to use their own user lists in campaigns. Context hints explain the conversational situations in which an Ad Group may be appropriate.
These two structures are not alternatives to each other.
For example, an e-commerce brand could create the following setup:
- Custom Audience: Customers who purchased within the last 180 days
- Ad Group: New-season women’s dress collection
- Context hints: “Users looking for elegant and comfortable women’s dresses for a summer event, graduation, or special occasion”
- Bid multiplier: 2x for the high-value customer segment
This structure uses both the brand’s own customer data and the conversational context within ChatGPT.
When writing Context Hints, it is more effective to describe the user’s need, decision criteria, and use case instead of creating a keyword list.
How should audiences be named?
A consistent audience naming convention makes campaign management easier. Standardized naming is particularly important when using multiple customer lists, lead lists, or category-based segments.
Example naming structure:
TR | Customers | Last 180 Days | Email | 2026-07
TR | Leads | Newsletter | Email | 2026-07
TR | Purchasers | High Value | Phone | 2026-07
This structure makes it possible to understand the audience’s country, type, scope, identifier format, and creation date at a glance.
Strategic recommendation: Avoid generic names such as “test,” “list 1,” or “customers.” As of July 2026, the audience name cannot be edited after creation, so the naming convention should be defined in advance.
Example audience structures for e-commerce brands
E-commerce brands can consider the following segments when using Custom Audiences in OpenAI Ads:
- Customers who purchased within the last 180 days
- Users who purchased within the last 30 days
- Newsletter subscribers
- Loyalty program members
- Customers with a high average order value
- Users interested in a specific category
- Users who submitted a lead during a campaign period
These segments should not all be used in the same way.
For example, users who purchased within the last 30 days may be excluded from certain acquisition campaigns. Higher bid multipliers may be applied to high-value customers within selected Ad Groups.
How can Optifeed support this process?
Optifeed is an AI-powered product feed optimization platform that helps e-commerce brands prepare their product data for advertising platforms and AI shopping experiences.
An audience strategy in OpenAI Ads involves more than uploading an audience file. Product titles, descriptions, prices, availability, categories, and product segments also affect the performance of the products promoted by the campaign.
Optifeed can support brands by:
- Segmenting products according to performance, availability, price, and category data
- Making product titles and descriptions easier to understand with AI Enrich
- Generating key product attributes and up to 30 product-related questions and answers
- Organizing and maintaining current price, availability, and variant information
- Creating product feed structures that comply with advertising channel requirements
- Making product data cleaner and easier to understand for AI shopping experiences
To explore product grouping based on performance signals in more detail, read AI-Powered Smart Product Segmentation.
To understand why product feed preparation is important for AI shopping, you can also read How to Prepare an AI Shopping-Ready Product Feed.
