The feature set that just shipped
OpenAI expanded the ChatGPT Ads toolset throughout August in ways that didn't get much attention individually but add up to a meaningful platform shift. According to Search Engine Land, the platform shipped oCPC (conversion-optimized cost-per-click) bidding in beta for product feed campaigns, a multi-product carousel format entering testing, dynamic URL parameter support for cleaner attribution, and custom audience targeting from hashed email and phone lists.
Taken together, this is ChatGPT Ads growing up fast. It went from a flat-CPC, single-format product to something that structurally resembles Meta or Google's mid-funnel toolkit — within roughly three months of self-serve launch in May.
What oCPC actually means
You still pay per click. But the system selects which clicks to chase based on their likelihood to convert, not just their likelihood to happen. OpenAI uses the selected conversion event — purchase, lead, registration — alongside ad quality, click probability, and conversion probability to bias delivery toward the clicks that matter.
This is how Meta's early bidding worked before it moved to fully automated CPA models. It's a middle step: you get optimization pressure toward outcomes while the cost basis stays predictable per click. For DTC brands used to chasing target ROAS, oCPC is the closest thing ChatGPT Ads has offered to performance marketing logic.
The catch is data. To rank clicks by conversion likelihood, the system needs to have seen enough conversions to build that probability estimate. If your ChatGPT Ads account has sent thirty clicks to your store and recorded two purchases, the oCPC signal is essentially random noise dressed up as optimization.
The learning phase isn't hypothetical
Google's Smart Bidding needs roughly 50 conversions per month to stabilize. Meta needs enough pixel signal to distinguish buyers from browsers across its user graph. ChatGPT Ads is newer, the user base is smaller, and purchase-intent signals are being built from scratch.
That's not a reason to avoid oCPC — it's a reason to start now. Every month you wait increases the gap between you and accounts that have been accumulating conversion history since May. The platform's learning models improve as they see more data. Accounts with six months of signal will get meaningfully better optimization than accounts starting cold when ChatGPT Ads eventually scales into a mainstream channel.
The practical issue is margin tolerance during learning. oCPC will bid inefficiently early on — the system doesn't know your customer profile yet. A 4–6 week learning phase with elevated or unpredictable CPCs is a real cost. Brands with thin margins and limited ad budgets should sequence this carefully: build CPC volume first, then flip to oCPC once the pixel has something to work with.
The carousel format changes the surface area
The multi-product carousel is still in testing per Search Engine Land, but the format matters for e-commerce brands more than the bidding change. A single-product ad forces ChatGPT to guess which product a given user wants before showing the ad. A carousel lets the user make that selection themselves — which is exactly the mechanic that made Meta's dynamic product ads so effective for catalog-heavy DTC brands.
If your catalog has meaningful SKU variety and your best customers tend to browse before buying, the carousel format is worth watching closely. It could turn ChatGPT Ads from a brand-discovery play into a catalog-conversion play, which is a very different budget conversation.
Who should activate oCPC this week
Activate oCPC now if you already have a live ChatGPT Ads CPC campaign, a verified Web pixel, and Automatic Advanced Matching turned on from the August 17 update. That combination means the system has some first-party data to start with.
Wait if you haven't run any ChatGPT Ads yet. Start with standard CPC to generate baseline click and conversion data first. Launching cold into oCPC is burning budget on a model that has nothing to learn from — the optimization pressure becomes random, not useful.
Either way, add dynamic URL parameters today. They let you automatically append campaign, ad group, and ad IDs to landing page URLs — no cost, takes ten minutes, and makes attribution in GA4 or your MMP dramatically cleaner when you eventually scale oCPC or add carousel formats.
If your ChatGPT Ads setup needs a full audit before you layer in conversion bidding, the free account audit at Gromerce takes about three minutes and shows you exactly where your signal gaps are.
ChatGPT Ads is building its ad stack faster than most expected. The platform won't replace Meta or Google this year — but the accounts doing the signal work now are the ones who'll use it effectively when it does.
Sources: Search Engine Land, ALM Corp, August 2026
