ChatGPT Ads Adds Conversions, Pacing, and MMPs to Chase Google

Any performance lead who has ever tried to justify a new ad channel to a CFO knows the two questions that decide the meeting: can you optimise on conversions, and can you measure app installs. Until this week, ChatGPT Ads answered "no" to both. That has now changed, and the shift arrived quietly in a product update rather than a keynote.

OpenAI's ad product just grew up several release cycles at once. The question for buyers is whether the plumbing underneath is production-grade or still a beta wearing a suit.

What Happened

OpenAI pushed a batch of upgrades to ChatGPT Ads covering optimisation, measurement, and budget control. As Campaign US reported, advertisers can now create conversion-optimised campaigns by selecting a Conversions objective. The system automatically optimises toward clicks more likely to generate conversions, while still charging on a CPC basis. That is a meaningful pricing detail: buyers get smarter delivery without moving to a CPA or oCPM billing model.

Daily budgets have shifted to an average daily budget model over a rolling seven-day window. Spend can fluctuate day to day but stays within overall limits. On top of that, automatic budget pacing now spreads spend more evenly throughout the day, closing the classic morning-blowout hole that eats junior media buyers alive.

Geographic exclusions arrived too, letting advertisers carve specific locations out of targeting. That is table stakes for regulated verticals, and its absence was frankly embarrassing until now.

On measurement, ChatGPT Ads launched integrations with AppsFlyer and Adjust so advertisers can measure app installs and in-app events. Web-side attribution gets a lift from Automatic Advanced Matching, which uses hashed customer data to improve conversion attribution and is toggled under Tools > Conversions > Data Source. The Bulk API now supports asynchronous bulk creation and updates for campaigns, ad groups, and ads. Product feed campaigns are also starting to show refreshed cards with pricing and star ratings. OpenAI is explicitly positioning the product against Google Ads and Meta Ads.

Technical Anatomy

Strip away the marketing gloss and this release is really four engineering decisions worth understanding.

First, the Conversions objective on a CPC billing base. That combination tells you the auction is still click-priced, but the ranking function has been extended with a conversion propensity model. Buyers get bid-time optimisation without OpenAI having to guarantee CPA outcomes. It also means attribution signal quality directly determines how well the model learns. Garbage pixel, garbage bidding. This is why Automatic Advanced Matching matters: hashed first-party identifiers (email, phone, likely normalised the same way Meta's Conversions API handles them) feed the matching layer and improve the closed loop.

Second, average daily budgets on a rolling seven-day window plus intraday pacing. That is a real infra change, not a UI flag. Pacing throughout the day requires a spend forecaster running against auction volume, decrementing budget in near-real-time across ad server nodes. Teams I've worked with at scale know this component is where bugs melt production: a stale budget cache and you overspend a client's monthly cap by lunchtime. The seven-day rolling average also means finance dashboards need to stop treating daily spend as a hard cap. Reporting logic has to change downstream.

Third, mobile measurement via AppsFlyer and Adjust. These are the two MMPs that own iGaming, fintech, and mobile-first commerce attribution. Integrating with both, rather than picking one, signals OpenAI wants app advertisers on day one, not year two. The install and in-app event postbacks will look similar to what Google and Meta already consume.

Fourth, the asynchronous Bulk API. Async bulk endpoints are what separate a demo API from something an agency can actually run at scale. If you have ever tried to push five thousand ad variants through a synchronous endpoint, you know why this matters. It also implies OpenAI is finally comfortable enough with backend throughput to expose queued writes. Google's Ads API has had this pattern for years.

My take: the ordering of these releases tells you the roadmap. Optimisation, measurement, bulk operations. That is the exact sequence any serious ad platform ships when it wants agency budgets, not indie experimenters.

Who Gets Burned

Google and Meta are the obvious names, but the near-term pain is more distributed than that. The uncomfortable read: mid-market performance agencies that built their moat on "we know the ChatGPT Ads quirks" just watched half those quirks disappear. When budget pacing and conversion optimisation are automated, the buyer's edge collapses toward creative and feed quality.

App-install shops in iGaming and fintech now have a third serious channel to test. That is genuinely disruptive. For years the mobile UA stack has been a duopoly with a long tail of underperforming networks. An OpenAI-owned surface with AppsFlyer and Adjust support, sitting inside a product with hundreds of millions of weekly users, is not a rounding-error experiment. Media plans get rewritten.

Attribution vendors themselves benefit but also get squeezed. AppsFlyer and Adjust become gatekeepers for another major channel, which is good for revenue but bad for negotiation use: OpenAI now knows exactly how much they need each vendor. Expect quiet re-pricing conversations within the next two quarters.

On the ad ops side, teams running Google Ads Editor style bulk workflows now need a second script path. If you support agencies, the Bulk API being asynchronous means your job queue architecture has to handle callbacks and status polling, not just fire-and-forget uploads. Production incidents I've seen in this pattern usually come from missing retry logic on transient 5xx responses. Build it in on day one.

Finally, ecommerce brands with product feeds catch a small tailwind. Refreshed cards with pricing and star ratings mean feed quality directly moves CTR again, exactly like it did when Google Shopping matured. If your feed hygiene is poor, you will be the case study.

Playbook for Performance Marketing

Concrete moves for the next two weeks, in priority order.

Turn on Automatic Advanced Matching before you touch anything else. Hashed first-party data feeding the conversion model is the single highest-use change here. Without it, the new Conversions objective is bidding half-blind.

Run a controlled test of the Conversions objective against your existing CPC campaigns. Same creative, same audience, same budget. Give it a full seven-day cycle to align with the new rolling budget model. Do not judge on day-three numbers, the pacing algorithm needs a week of learning data.

If you are a mobile advertiser, wire up AppsFlyer or Adjust this sprint. Getting install and in-app event postbacks flowing early means you own the incrementality data before your competitors do. That is worth more than any bid strategy tweak.

Rebuild your budget monitoring. Daily hard-cap alerts will fire false positives under the new average daily model. Move your finance dashboards to seven-day rolling views or you will have some awkward Monday morning calls.

Refactor your bulk upload scripts to handle async responses. Job IDs, status polling, exponential backoff on retries. If you skip this, you will discover it during your first thousand-ad push, at the worst possible moment.

Ship geographic exclusions immediately for any regulated vertical (gambling, crypto, financial services). This was a compliance gap and it just closed. Use it.

Key Takeaways

• ChatGPT Ads now supports a Conversions objective on CPC billing, average daily budgets over a rolling seven-day window, and intraday automatic pacing.
• AppsFlyer and Adjust integrations bring app install and in-app event measurement, opening the channel to mobile-first iGaming and fintech advertisers.
• Automatic Advanced Matching under Tools > Conversions > Data Source is the highest-use first switch to flip, because it feeds the new bidding model.
• The Bulk API's async support signals OpenAI is targeting agency-scale workflows, not just self-serve experimenters.
• Refreshed product cards with pricing and star ratings make feed hygiene a direct performance lever again for ecommerce brands.