OpenAI emailed ChatGPT Ads Manager Beta participants on May 22 with a short product update. Four additions: a daily budget option for new campaigns, geo-targeting at the state, DMA, and ZIP-code level, aggregate reporting across campaign and ad-group layers, and early testing of algorithm-selected call-to-action buttons. Small announcement. Significant trajectory.
What the geo update actually unlocks
Since the beta launched, every ChatGPT ad campaign targeted the entire United States. No exceptions. That excluded any brand without national reach and made the platform irrelevant for regional e-commerce, local retail, and service businesses with geographic limits.
State and DMA-level targeting changes that. A furniture brand that ships only to the Northeast can now run a ChatGPT Ads test without spending against California traffic that can't convert. A regional grocery chain can run against their actual DMAs. More importantly, DMA targeting maps to the geography TV media buyers already work in, which means brands with TV budgets can benchmark ChatGPT Ads directly against their linear spend in the same market.
ZIP-code targeting at launch is ahead of where both Google and Meta started when they opened their platforms. OpenAI is not building incrementally.
Daily budgets close a basic workflow gap
Lifetime budgets work for flight-specific campaigns. They don't work for always-on accounts, which is how most performance teams operate. Without daily budget options, ChatGPT Ads was structurally incompatible with how performance media is actually managed.
Daily budgets mean you can run this channel like any other line item: set a cap, watch it pace, adjust weekly. That's not a minor convenience addition. It's the difference between a platform you can slot into a media plan and one you have to account for separately every month.
Dynamic CTAs and what they tell you about the platform's direction
A small subset of ChatGPT ads will now show algorithm-selected call-to-action buttons. OpenAI's system reads the ad creative and the destination landing page, then picks the CTA it thinks fits best. You don't choose it.
This is the same mechanism Google used when it introduced Final URL Expansion and asset-based creative generation. The pattern is consistent: the platform uses your content as raw material and optimizes around it. Your landing page becomes the creative brief.
If your product pages are thin, the CTA output will be generic. If your pages cover specific use cases and buyer questions, the algorithm has something to work with. This is now the third ad surface in a single month where landing page depth determines the ad quality ceiling — after Conversational Discovery Ads and AI Max for Google.
The feature buildout pattern
In roughly 10 weeks since the ChatGPT Ads beta opened:
CPMs dropped from $60 to $25. CPC bidding was added at $3–5 per click. Ads Manager opened to all US businesses. Conversion pixel shipped. Product feed automation landed for up to one million SKUs. Now: geo-targeting, daily budgets, reporting layers, and dynamic CTAs.
That's not gradual iteration. It's an accelerated checklist of features that separate "branding experiment" from "performance platform." The logical next additions are audience targeting (custom lists, lookalikes, retargeting), frequency controls, and conversion-optimized bidding. When those ship, the "I'll test when it's ready" stance runs out of runway.
What to do with this right now
If you're outside the beta, a minimum spend requirement likely still keeps this out of reach for most accounts. Watch the trajectory but don't move budget yet.
If you're in the beta, geo controls and daily budgets make a real comparative test possible for the first time. Pick one state or DMA where your brand has the strongest market position, run 30 days with a controlled daily budget, and put the results next to your Google AI Mode and Meta performance in that same geography. That data point is worth collecting early.
On dynamic CTAs: don't override them for the first two to three weeks. Let the algorithm select, then observe what it's generating. If the outputs are generic ("Shop Now," "Learn More"), your landing pages are the problem — not the platform.
One thing worth starting now regardless of beta access: audit the product pages you'd route ChatGPT Ads traffic to. Pages built for keyword density will produce weak AI-generated ad outputs. Pages that answer real customer questions will perform differently on every AI ad surface coming out this year.
If you want a clearer picture of how your accounts are positioned across these new AI-native surfaces, the free audit at Gromerce surfaces the gaps in a few minutes.
OpenAI is building a real ad platform inside ChatGPT. The pace of the feature additions means the window to establish early baselines is shorter than it looks.
Sources: Search Engine Land, Performance Marketing World, PPC.land, Marketing4eCommerce, May 2026
