ChatGPT Ads 2026: Budget Rules and Conversion Changes

What Changed on July 27, 2026, and What Changed Later?

The July announcement brought several advertiser-facing changes together. Search Engine Roundtable reported the changes using communications sent to ChatGPT advertisers, including average daily budgets, more automated spend pacing, conversion-optimised campaigns, location exclusions, mobile attribution integrations, matching improvements, bulk operations and a refreshed product-card format.

These developments should not be treated as one simultaneous general-availability launch. Some were policy changes affecting existing campaigns. Others were new campaign options, integrations or formats whose availability depended on the advertising account or rollout stage.

There is another important distinction. OpenAI’s Help Center has continued evolving after the July announcement. A feature described as unavailable or limited in an earlier report may have a different configuration in the current Ads Manager Beta documentation.

For an advertiser evaluating the platform, the useful question is therefore not simply what OpenAI announced in July. It is which operational rules apply now, which settings require a new campaign, and which features remain restricted.

A Seven-Day Average Is Not a Daily Spending Ceiling

Under the updated daily budget system, advertisers choose an average amount that guides delivery across a seven-day budget week. OpenAI’s current documentation defines that week as Sunday through Saturday in the advertising account’s timezone.

The selected daily budget is not a guarantee of identical expenditure each day. Delivery can rise when advertising opportunities are available and fall when demand or eligible inventory is lower.

OpenAI documents two spending controls:

• Daily limit: billed media spend may reach up to twice the selected daily budget on an individual day.
• Seven-day limit: billed media spend cannot exceed seven times the selected daily budget over the applicable seven-day period, subject to the documented rules for budget changes.

For example, a campaign with a $100 daily budget can have a $700 spending limit across a full seven-day period. An individual day can reach $200 without necessarily violating the campaign’s budget rules.

That does not mean advertisers should expect the campaign to spend $200 on busy days or the full $700 every week. OpenAI’s pacing system adjusts delivery, and spending may remain below the selected budget.

The rules also become more involved when an advertiser changes the budget during a day or partway through the week. OpenAI applies the highest active daily amount when determining that day’s maximum and prorates the applicable seven-day limit following certain changes.

Existing campaigns moved to the average daily budget model automatically. The original July announcement did not require advertisers to recreate their campaigns or increase their selected budget amounts.

The consequence for campaign reviews is straightforward: a single day’s above-average spend is not sufficient evidence of overspending. Advertisers need to compare delivered spend with the applicable daily and seven-day limits, using the account’s timezone and budget-change history.

See OpenAI’s Daily Budgets documentation for the current calculation rules.

Conversion Optimisation Now Has Two Billing Options

The original July reporting focused on conversion-optimised cost-per-click campaigns, usually described as oCPC. These campaigns use conversion signals to prioritise clicks that are more likely to lead to a selected business outcome, while charging for valid clicks.

That remains one supported model, but it no longer describes the complete conversion optimisation offering.

OpenAI’s current documentation also describes conversion optimisation with impression billing, or oCPM. Both options aim to optimise delivery toward the selected conversion event, but they charge advertisers differently.

Neither option is a cost-per-acquisition contract. Advertisers are not billed only after a purchase, registration or lead has been completed.

OpenAI requires a supported standard conversion event for the conversion optimisation objective. Custom conversion events are not currently supported for selecting that optimisation goal.

Advertisers also need to distinguish changing an existing campaign from creating a replacement. OpenAI documents cloning an existing campaign and selecting the conversion optimisation objective for the cloned version. The desired conversion event cannot be changed after campaign creation; a different event requires another campaign.

This makes campaign structure and measurement configuration more consequential than the label attached to a bidding model. A conversion campaign cannot reliably optimise toward a business outcome if the relevant event is missing, incorrectly named or too infrequent to provide useful signals.

The current setup requirements are explained in OpenAI’s Conversion-Optimized Campaigns guide.

Why the Billing Difference Matters

Click billing and impression billing create different ways to evaluate costs. An advertiser using oCPC should examine clicks, conversion volume and cost per conversion together. An advertiser using oCPM should also understand impression exposure and, where available, the difference between conversions attributed to clicks and those attributed to views.

A lower reported cost per conversion does not automatically establish that one billing method is superior. Attribution windows, event quality and differences in the audience reached can affect the comparison.

It is also premature to assume that ChatGPT Ads will consistently outperform established paid search campaigns. The reported optimisation mechanism describes how delivery is selected, not a verified return on advertising spend.

Mobile Attribution Has Expanded Beyond the Original Two Partners

The July 2026 announcement highlighted AppsFlyer and Adjust integrations for measuring app installs and selected in-app activity attributed to ChatGPT advertising.

OpenAI’s current mobile measurement documentation identifies a broader supported set: Adjust, Airbridge, AppsFlyer, Branch, Kochava, Singular and Tenjin.

This does not mean every partner supports the same implementation or reporting options. Setup, event mapping and eligible measurement capabilities vary. The advertiser must still establish the appropriate relationship between its Ads Manager account, mobile measurement partner and conversion events.

OpenAI’s documented setup commonly involves an Ads Manager Pixel ID, a Conversions API key, partner permissions, selected event names and a tracking link or other supported integration method.

Measurement partners can help connect advertising activity to app outcomes, but advertisers should not treat all reported events as immediately available or directly comparable across platforms.

For example, a discrepancy between an app analytics platform and Ads Manager may reflect different event mappings, attribution rules, processing delays or measurement eligibility. The first response should be to inspect the implementation rather than assume one platform is wrong.

The current list and partner-specific instructions are in OpenAI’s Mobile Measurement Partner Integrations guide.

Automatic Advanced Matching Changes Measurement, Not Consent Obligations

Automatic Advanced Matching, or AAM, was part of the July discussion because it offered a way to associate more eligible website conversions with advertising interactions.

The earlier rollout communication described default enablement for newly created Web pixels and a planned August 17, 2026 change for existing pixels. That historical date should no longer be presented as an upcoming action.

Current OpenAI documentation describes the mechanism rather than relying on that historical activation timetable. The OpenAI Pixel can detect supported customer information from recognised website forms and other supported sources, normalise it and hash it in the browser using SHA-256 before including it with conversion events.

OpenAI states that raw customer information is not transmitted through this automatic matching process. Advertisers should nevertheless inspect their own current measurement settings rather than assume the feature is enabled or disabled based solely on when a pixel was created.

Hashing is not a substitute for the lawful collection and use of personal information. Website owners still need to provide appropriate information and obtain any legally required consent before sharing eligible conversion data.

The practical measurement question is whether the advertiser’s conversion events, click references and permitted matching signals accurately represent the desired outcome.

OpenAI also documents using the Pixel and Conversions API together. If both methods report the same conversion, they should use the same event ID so that the event can be deduplicated.

See OpenAI’s Conversion Measurement documentation for the current process and controls.

The Bulk API Is Useful, but Access Is Still Restricted

OpenAI’s Bulk API allows advertisers to create or update campaigns, ad groups and ads through asynchronous jobs rather than processing every operation individually.

It remains a limited-preview capability enabled at the advertising-account level. Its presence in public developer documentation does not mean every account can use it.

OpenAI currently documents up to 1,000 operations in a single bulk job. A job can contain related operations, such as creating a campaign and then creating associated ad groups and ads.

Advertisers using bulk operations need to examine job results, not merely confirm that a request was accepted. Some operations may succeed while others fail or are skipped. An HTTP response accepting the job does not establish that every requested campaign change has completed.

The documentation also supports validation-only requests and idempotency controls, which are particularly relevant when integrations need to retry operations without accidentally duplicating campaign objects.

Accounts without Bulk API access may receive a 404 response from the relevant endpoint. Before investing development resources, teams should confirm eligibility with their OpenAI account contact.

OpenAI’s Bulk API documentation describes the supported operations, access limitations and processing behaviour.

There is a separate distinction between the limited-preview Bulk API and bulk campaign workflows available through the Ads Manager interface. OpenAI documents CSV-based bulk creation for supported campaign types, but that interface does not currently support creating product-feed campaigns through CSV upload.

Geo Exclusions and Product Cards Require Account-Level Confirmation

Search Engine Roundtable’s July report also described geographic exclusions and refreshed product-feed ad cards in advertiser communications.

Geo exclusions were presented as a way to remove specified locations from campaign delivery. This matters for advertisers operating in markets with different shipping coverage, product availability, service areas or legal requirements.

The refreshed product-card format was reported as including price information and star ratings, with rollout beginning around July 27, 2026.

These were reported product changes, not independently verified evidence that every account received the same format on that date.

The distinction is especially important for product-feed advertisers. A format preview may demonstrate how an advertisement could appear, but it does not confirm that the account is eligible to serve it or that it will improve commercial outcomes.

Advertisers should check their own available targeting controls, product-feed requirements and ad previews before relying on a new presentation format.

The original July developments were reported in Search Engine Roundtable’s coverage of ChatGPT Ads Manager changes.

What These Changes Mean for an Advertiser’s Operating Decisions

The most consequential development is not the number of new controls. ChatGPT Ads is becoming a platform where budgeting, delivery optimisation and measurement are increasingly interconnected.

That makes it less useful to evaluate the service solely through its original CPC announcement. MOCOBIN’s earlier report on ChatGPT’s initial CPC bidding model covers that earlier stage. The newer question is how advertisers should operate a campaign once conversion objectives, variable daily spending and more extensive attribution tools are available.

Three decisions deserve particular attention.

First, Separate Budget Limits From Delivery Targets

A selected daily amount is now an average target with defined spending limits. Advertisers who need a strict maximum across an entire campaign should evaluate the campaign-total budget option instead of assuming a daily setting provides a hard same-day ceiling.

Short-term pacing decisions should account for the platform’s reporting delays. OpenAI says impressions, clicks and CTR generally refresh approximately every 15 minutes, while spend data may be delayed by seven to eight hours. A temporary zero in the spend report is not evidence that no charges have accrued.

Second, Verify the Conversion Event Before Comparing Bids

The apparent sophistication of a conversion optimisation model cannot compensate for an incorrectly configured measurement system.

Before changing an objective or billing method, advertisers should verify the chosen standard event, test whether eligible events are received and check whether reporting reflects the same business outcome used in their internal analytics.

The comparison is not simply oCPC versus oCPM. It is a comparison of campaigns with different billing exposure, attribution rules and potentially different conversion signals.

Third, Treat New Automation as an Operational Responsibility

Bulk campaign tools can reduce repetitive work. They can also increase the scale of an error. Incorrect budgets, destination URLs or targeting settings become more consequential when an automated job changes hundreds of objects.

Validation, permissions, paused initial states and a process for reviewing partial failures should be part of an integration from the beginning.

These operational differences also help explain why ChatGPT advertising should not be treated as an extension of organic search visibility. Paid campaign delivery depends on eligibility, account controls, bidding and available budget. Organic SEO depends on a different system of discovery, indexing, relevance and ranking. MOCOBIN’s SEO fundamentals guide explains that distinction.

What Is Confirmed, and What Still Needs Evidence?

OpenAI has documented the budget rules, conversion billing options, measurement integrations and Bulk API limitations discussed here. These are product capabilities and policies, not independent evidence of commercial effectiveness.

Several questions still require account-specific or longer-term evidence:

• Whether oCPC or oCPM produces better business outcomes for a particular advertiser after measurement differences are accounted for.
• How consistently individual accounts receive newly introduced formats and targeting controls.
• Whether matching improvements materially change attributed conversion volume without introducing reporting misunderstandings.
• How much operational time bulk workflows save after validation, review and error handling are included.
• Whether ChatGPT Ads can justify additional spending compared with established advertising channels.

None of those outcomes can be established from feature announcements alone.

For advertisers considering a larger commitment, a controlled test with clearly defined conversion events and a separate budget is more defensible than extrapolating from another company’s results or from the existence of a new bidding model.

As Ads Manager Beta develops, its documentation is likely to continue changing. The July 2026 announcement remains an important point in the platform’s development, but current account settings and updated official guidance must govern actual campaign decisions.