On July 9, 2026, OpenAI sent an email to ChatGPT Ads advertisers announcing five changes at once in Ads Manager: custom audiences using first-party lists, a new account Overview tab, automatically suggested ad drafts, a redesigned ad card format, and campaign expansion to Japan and South Korea. This is not an isolated change: it is the culmination of three weeks during which OpenAI activated CPA bidding (June 5), published the terms for its audience and creative tools (June 17), launched an AI-generated ads feature (June), and previewed custom audiences before formally announcing them (July 1–7). If you manage ChatGPT Ads accounts, this is the week to bring your setup up to date.
This article explains exactly what changed, how custom audiences work, why “suggested drafts” are not the same as AI-generated ads (and why that distinction matters), and what you should review in your account before continuing to scale your budget on the platform.
What Did OpenAI Announce for ChatGPT Ads on July 9, 2026?
OpenAI announced five simultaneous updates to Ads Manager: custom audiences, an Overview tab, suggested ad drafts, a revamped ad card format, and campaign availability in Japan and South Korea. The announcement came directly to advertisers by email rather than as a press release: public coverage was pieced together from screenshots shared by PPC consultants in the days leading up to the announcement, before OpenAI officially confirmed it.
The pace of this update is the real story behind the story. In less than three months, ChatGPT Ads went from a closed pilot with a minimum spend of $200,000 to a self-service Ads Manager with no minimum spend, CPC bidding, a conversions API, pixel measurement, CPA bidding, and now first-party audiences. For a new advertising platform, that is an unusually rapid pace of maturation, and it means that what you write about ChatGPT Ads today could become outdated in weeks rather than years.
If you want the full picture of what ChatGPT Ads is, how it compares with Google and Meta, and its history since the February pilot, we already covered that foundation in ChatGPT Ads 2026: What Every Digital Marketing Agency Needs to Know. Here, we focus specifically on what changed this week.
What Are Custom Audiences and How Do They Work?
Custom audiences allow you to upload your own customer or prospect lists (by email address or phone number) to include, exclude, or adjust the bid for people who see your ads on ChatGPT. It is the same first-party data targeting logic you already use in Google Ads and Meta Ads, now available in a conversational environment that, until just a few weeks ago, was defined primarily by search intent.
In practice, there are three available uses. You can use an audience as a campaign inclusion, limiting eligibility to known customers or qualified prospects. You can use it as a campaign exclusion to avoid showing ads to recent buyers or audiences that should not see them. And you can apply a bid multiplier at the ad group level, ranging from 0.1x to 10x, to bid more aggressively for high-value customers or more selectively for lower-priority audiences. If a user belongs to more than one audience with a multiplier, the highest applicable multiplier is used.
The real limitation is the audience-size threshold: each audience needs at least 25,000 matched users, and OpenAI recommends 100,000 or more. Ads Manager does not display individual users or allow you to select specific people, so you work with the aggregated segment rather than the raw list. For an SMB with a customer base of a few thousand people, this feature simply does not apply yet; for a brand or agency with larger lists, this is the first time ChatGPT Ads feels comparable to the first-party data targeting you already know from other platforms.
What Are Suggested Ad Drafts, and Why Are They NOT AI-Generated?
Suggested ad drafts are a feature that pre-populates an ad draft (image, headline, and description) using existing metadata from your website, rather than content generated by artificial intelligence. This distinction is the part most likely to cause confusion because it sounds almost identical to another feature OpenAI launched in June: “generated” ads that actually use AI to create new variations.
The practical difference: a suggested draft takes information that already exists on your page (the title, description, and image you have already published) and assembles a candidate ad for you to review, edit, and assign to a campaign or ad group. It does not invent new copy or new images; it simply assembles your existing content into the ad format. It appears when OpenAI detects that your campaign could benefit from greater content coverage to optimize delivery, under the “Add new ad” option within the campaign view.
This feature deserves to be understood on its own: if you see “Add new ad” with a pre-filled draft based on your website, that is reordered metadata, not new AI-generated creative. Review it as you would any other draft, but understand that its source is your own content rather than generation from scratch.
How Is This Different From the “Generated” Ads Released in June, and Why Does It Matter?
In June, OpenAI separately launched a feature that does use AI to generate new ad variations within an existing campaign, under the same “Add new ad” menu but labeled “ads generated for you.” There, advertisers can allow ChatGPT to create the ad and then review, edit, and approve it before it goes live, in an approval workflow similar to what we saw in Meta’s creative solution launched at Cannes.
This AI-generation feature has not been friction-free. A PPC consultant specializing in the legal sector documented a specific issue in early July: when an advertiser accepts one of the AI-suggested ads, the platform removes all ads from that ad group and sends them back for review, an unexpected behavior that interrupts active campaigns without warning. Neither feature (AI-generated ads nor metadata-based drafts) has been formally named by OpenAI yet, so both are still evolving.
The pattern here is repeated throughout the 2026 agentic advertising industry: AI is generating or assembling creative faster than a team can manually review it, while platform governance and stability are lagging behind technical capabilities. It is the same pattern we documented when analyzing how AI is already managing campaigns with increasing autonomy in TikTok Ads: the feature's speed currently exceeds the platform's operational maturity. The practical conclusion is the same in both cases: test these features in a low-risk campaign before trusting them with your primary account.
What Else Changed: Overview, Ad Card, and Expansion to Japan and South Korea?
In addition to audiences and drafts, OpenAI added an Overview tab to Ads Manager where you can monitor account health, review recommended tasks to improve campaign performance, and analyze your key metrics in a larger, more flexible trend chart. It is the kind of dashboard you would already expect in Google Ads or Meta Ads Manager, and its arrival confirms that OpenAI is building the platform using the same visual vocabulary as the rest of the industry.
OpenAI has also begun rolling out a more compact and easier-to-read static ad card format, with larger visual elements, on web and mobile (a change that actually launched in late June but was included in OpenAI's July 9 summary). Geographically, campaigns can now target users in Japan and South Korea, expanding reach for advertisers doing business in those markets, as part of an international expansion OpenAI had previewed in May when it announced plans to bring the advertising pilot to the UK, Mexico, Brazil, Japan, and South Korea.
None of these three changes requires an urgent decision on your part, unlike custom audiences or AI-generated ads. They are platform improvements that you simply absorb as you log in: worth knowing about, but they do not change your strategy in the same way.
Should My Brand Start Advertising on ChatGPT Ads Right Now?
It depends on how well-developed your conversion funnel is and how comfortable you are with a platform that is still changing week by week. The barrier to entry is low: a self-service Ads Manager with no minimum spend, CPC bidding at around $3–$5 as recommended by OpenAI, and both a conversions API and pixel for measuring actual results instead of guessing. That makes testing the platform with a limited budget a relatively low-risk decision.
What should be treated cautiously is the performance reported so far. Some advertisers have reported click-through rates as low as 0.91%, compared with a benchmark of around 6.4% in Google Search, but the 0.91% figure is data reported by advertisers through the press, not a benchmark published by OpenAI, and it compares fundamentally different surfaces (a conversational response versus a search results page). Treat it as a directional signal, not a number on which to base your budget.
If your industry falls into regulated categories, there is an additional layer of due diligence. OpenAI has contradictory documentation regarding which categories it allows: legal services and healthcare appear both on its list of restricted categories and its list of prohibited categories, and yet we know that, as of July 2026, legal-services ads are indeed being served on the platform. If you operate in a regulated vertical, do not assume anything based on public documentation: verify the actual status of your category before launching.
This is where it is worth delegating ongoing monitoring to someone who tracks these changes for you. At JP Director, we monitor this type of platform update for our clients in the same way we follow Meta, Google, and TikTok: not to chase every new feature the day it launches, but to make informed decisions about which ones are worth adopting now and which should be allowed to mature before committing real budget.
To illustrate this with a hypothetical case (not an actual campaign, but a typical scenario), imagine an account testing ChatGPT Ads with a limited daily budget, uploading a custom audience of existing customers as an exclusion (to avoid spending money on people who have already purchased), and installing the conversion pixel from day one. That order of operations—exclusion before inclusion, measurement before scaling—is what separates a properly instrumented test from speculative spending on a platform that is still taking shape.
Frequently Asked Questions
Do Suggested Ad Drafts in ChatGPT Ads Use Artificial Intelligence?
No. Suggested ad drafts use metadata that already exists on your website (title, description, and image) to pre-fill an ad draft that you review, edit, and assign. They do not generate new copy or images with AI. That feature is different: it is unofficially called “ads generated for you,” was launched in June 2026, and does use AI to create new ad variations that the advertiser can approve.
What Is the Minimum List Size to Use Custom Audiences in ChatGPT Ads?
The minimum is 25,000 matched users per audience, and OpenAI recommends at least 100,000 for optimal performance. You can create the audience by uploading email addresses or phone numbers and use it to include or exclude campaigns or adjust bidding with a multiplier between 0.1x and 10x at the ad-group level. For businesses with small customer lists, this feature is not yet practical.
Is It True That ChatGPT Ads Has a Much Lower Click-Through Rate Than Google Search?
There are advertiser reports citing a click-through rate as low as 0.91% on ChatGPT Ads, compared with a benchmark of around 6.4% in Google Search, but that figure comes from advertiser reports relayed by specialized media rather than an official benchmark published by OpenAI. The comparison also mixes two different surfaces (a conversational response versus a search results page), so it should be treated as a directional signal rather than a solid basis for budget projections.
Can My Agency Create a ChatGPT Ads Account on Behalf of My Brand?
Not according to OpenAI's current policy: the platform explicitly states that agencies should not create advertising accounts on behalf of their clients. The client must create their own advertiser account and invite agency users to manage it from there. This is an operational detail worth resolving before signing any scope of work that includes ChatGPT Ads management.
