ChatGPT Ads Just Got Custom Audiences: First-Party Targeting Arrives

OpenAI quietly launched Custom Audiences for ChatGPT Ads in early July, before any formal announcement, and it's the clearest sign yet that ChatGPT is building toward the same performance marketing toolkit advertisers already use on Google and Meta. You can now upload your own list of customers or prospects using email addresses or phone numbers, and use that list to include or exclude audiences at the campaign level, or apply bid adjustments for matched users. First-party data targeting has arrived on a platform that, six months ago, only ran contextual ads matched to conversation topics.

How Custom Audiences Actually Work

Advertisers upload a list built from their own customer or prospect data, matched using email addresses or phone numbers, the same hashed-match approach Google and Meta have used for years with Customer Match and Custom Audiences. Once matched, that audience can be applied three ways: included so a campaign only reaches people on the list, excluded so a campaign avoids people already converted or already customers, or used to apply bid adjustments so matched users get different bidding treatment than the general population. OpenAI's terms explicitly rule out list rentals, third-party broker data, and scraped leads, only first-party data you collected directly qualifies.

Why the 25,000-User Minimum Matters

A confirmed floor of 25,000 matched users per list immediately filters out who this feature is actually for. A small local business with a few hundred email subscribers can't use this yet, the minimum assumes a customer or prospect database at a scale closer to mid-market and enterprise advertisers. That's consistent with where ChatGPT Ads has focused its rollout generally, and it's worth remembering if you're evaluating whether this feature is relevant to your account size right now versus something to revisit once your list grows.

What This Means for Your Channel Strategy

• If you already run Customer Match on Google or Custom Audiences on Meta, the same list works here in principle. The matching mechanics are similar enough that a business with a mature first-party data practice can extend targeting to ChatGPT without building anything new, just meeting the 25,000-user threshold.
• Exclusion targeting is the lower-risk, higher-value starting point. Using your customer list to exclude existing buyers from acquisition campaigns is a safer first test than inclusion targeting, since it protects budget from being spent reaching people who already converted, regardless of how well ChatGPT's inclusion matching performs.
• This is a meaningful signal about where ChatGPT Ads is headed, not yet a reason to shift budget. A performance marketing feature set this close to what Google and Meta already offer suggests OpenAI is building for scale, but a brand-new audience targeting feature with limited public performance data is a test budget decision, not a reallocation decision.

A new ad platform adding first-party audience targeting this early in its life is worth testing before it becomes crowded and expensive, but worth testing carefully, with a list and a budget you can afford to learn from. If you're weighing whether ChatGPT Ads' custom audiences fit your acquisition or retention strategy, our paid media team can help you structure a contained test using data you're already collecting.