ChatGPT Ads is not a product that becomes identical for every advertiser on the same day. In September 2026, it combines a documented technical foundation, tools that depend on account type, capabilities moving through a gradual rollout, and features announced for a later stage. This distinction matters: an item mentioned in an advertiser email or changelog is not automatically available in every account, and it should not become a campaign dependency before it is verified.
This article is an operational snapshot dated September 16, 2026. It does not repeat the broad overview in our ChatGPT Ads guide for businesses in Romania. Instead, it answers a narrower question: what can an advertiser rely on today, and what should still be treated as conditional access, staged rollout, or product direction?
A practical rule: a feature is genuinely available to a campaign only when it appears in the correct account, can be configured, passes the required reviews, and produces data in reporting. Public documentation proves that a product capability exists; it does not prove that every advertiser is eligible for it.
Why availability is not a simple yes-or-no question
A new advertising platform has several layers of access. A company may have an Ads account without product-feed access. It may see a conversion objective but lack a healthy Pixel or enough events. It may connect a workspace tool while delivery remains blocked by brand review, creative review, billing configuration, targeting restrictions, or the amount of eligible inventory in the selected market.
For that reason, this review uses four maturity states rather than a single “launched” label. Each state can apply independently to a specific capability:
| Status | What it means in practice | How an advertiser should treat it |
|---|---|---|
| Live and documented | It appears in current documentation and eligible accounts can use it in the described configuration | It may enter the plan after account and policy checks |
| Live but conditional | The feature exists, but access depends on account approval, billing, campaign type, audience size, or another requirement | Plan around it only after direct account confirmation |
| Rolling out | OpenAI is expanding access gradually, so two accounts may show different controls during the same period | Prepare the process without promising an activation date |
| Announced | The direction has been communicated, but operational documentation or broad access is not yet sufficient | Do not build forecasts or critical dependencies around it |
There is also a fifth category worth naming plainly: unconfirmed claims. Screenshots from another account, community posts, and promotional wording may suggest a capability that does not appear in public documentation or in the advertiser's own account. In that situation, the accurate answer is “we cannot plan around it yet,” not “it will probably work.”
The active and documented core
The Advertiser API documentation already describes the fundamental campaign structure: ad account, API key, campaign, ad group, ad, targeting, budget, and reporting. It documents objectives for impressions, clicks, and conversions, together with create, read, and update operations. That means ChatGPT Ads has moved beyond a concept demonstration, but it does not remove eligibility checks or delivery constraints.
Campaign management and review states
Campaigns, ad groups, and ads each have their own status. A campaign can be marked active while a child ad remains under review or has been rejected. Schedule, budget, account spending controls, and targeting can stop delivery even when the top-level object says “active.” Activation should therefore never be confused with effective publication or with immediate impressions.
A reliable workflow checks the entire hierarchy: account, campaign, ad group, ad, and every resource used by that ad. Product-feed campaigns add feed and product states. Conversion optimization adds the selected event and the health of the measurement pipeline. This is why a controlled delivery test tells the team much more than a screenshot of one active toggle.
Measurement Pixel and Conversions API
OpenAI documents the Measurement Pixel for browser events and Conversions API for events sent from a server. The two can work together when a shared event identifier is used for deduplication. Standard events are supported, and eligible matching fields can accompany value and currency information, subject to consent and data-protection requirements.
Reporting includes click-through conversions and, in eligible situations, view-through conversions. At the time of this review, view-through attribution uses a one-day window and an eligible click takes precedence. The critical detail is that view-through conversions remain supplemental reporting. Current CPA, post-click conversion rate, and conversion optimization are still based on outcomes that follow a click.
Maximize Results bidding
The maximize_clicks and maximize_conversions strategies are documented for standard and product-feed campaigns. Both require a daily budget and use clicks as the billing event. Conversion optimization requires exactly one active standard event, selected explicitly. The feature is live, but it cannot repair a conversion signal that is duplicated, too early in the customer journey, or disconnected from commercial value.
Our practical guide to ChatGPT Ads audiences, conversions, carousels, and budgets explains the compatibility rules in detail. For this availability map, the key distinction is simpler: Maximize Results is active, while the announced model expected to optimize with view-through outcomes is a separate, later capability.
Custom audiences and platform targeting
Custom-audience management is active in the current documentation. Members can be added, removed, replaced, or merged without continuously recreating the audience. Supported inputs include identifiers such as email, phone, and Google Advertising ID in accepted formats. Inclusion and bid multipliers require at least 25,000 matched users. A ready audience used only for exclusion currently has no documented minimum match count.
Platform targeting covers iOS apps, Android apps, and web. The September 10, 2026 update added more granular web options: desktop web, iOS web, and Android web. Broad web targeting remains available. Geographic eligibility and platform restrictions need to be reviewed together, because selecting a country does not guarantee enough eligible inventory or audience volume to deliver consistently.
Live capabilities that are not equal across accounts
This is where most planning errors begin. Documentation may be complete while access remains limited to specific accounts or commercial configurations. An agency should not present these options as standard before inspecting the advertiser's actual account.
| Capability | Status on September 16, 2026 | Condition to verify |
|---|---|---|
| Product feeds | Documented and usable by accounts with access | The account needs Product Feed API access, and product pages and images must be publicly accessible |
| Carousel card metrics | Live in reporting | The campaign must serve a multi-product format and generate product-level data |
| Daily account spending limit | Added on September 9, 2026 | Documented for eligible postpaid invoice accounts |
| Spending-limit windows | Available only to some accounts | The option must exist in that account's configuration; broad access should not be assumed |
| Ads Manager plugin in ChatGPT or Codex | Access depends on account, workspace, role, and rollout | The plugin must be visible, connectable, and authorized for the correct ad account |
A product feed is an entitlement, not merely a file
A technically valid catalog is not enough. The account needs feed access, and every product needs a publicly reachable landing page and image. Products can be rejected or remain eligible without receiving delivery. Product- and card-level reporting helps diagnose the outcome, but only appears when the campaign format and underlying data exist. If the endpoint or account option is unavailable, the documented path is to contact the account team, not to imitate the feature with an unsupported workflow.
Spending controls depend on billing
The daily account spending limit added to the changelog on September 9 explicitly applies to eligible postpaid invoice accounts. It is separate from campaign budget. A campaign can still have room within its own limit and stop because the account-level cap has been reached. Spending-limit windows, by contrast, are described as available only to some accounts. Finance and media teams need to account for these differences when several campaigns share the same ad account.
The plugin can be present without every action being available
The Ads Manager plugin supports conversational work with campaign structure, analysis, and proposed changes. Its presence in a workspace does not automatically grant access to every ad account, role, or product feature. Connection, permission scope, preview, and confirmation remain essential. A response produced in conversation is an operating aid, not evidence that the creative has passed review or can deliver in Romania.
Recent additions that can be used now
The update history is valuable because it separates changes that have entered the API from broader product announcements. Granular web targeting arrived on September 10, 2026, and the daily account spending limit for eligible billing configurations arrived on September 9. On August 25, audience operations were expanded to cover more flexible member updates and the use of small or empty audiences for exclusion.
Other documented additions include conversion-optimized bidding introduced in June, reporting breakdowns by product, country, and device, and expanded visibility into products with zero impressions. These should not be presented as future features. Their presence in the API, however, does not guarantee immediate volume or identical interface rollout in every account.
| Update date | Capability | Accurate interpretation |
|---|---|---|
| September 10, 2026 | Separate targeting for desktop web, iOS web, and Android web | The options are documented; segmentation still needs enough volume and useful results |
| September 9, 2026 | Daily ad-account spending limit | Available to eligible postpaid invoice accounts, not every billing setup |
| August 25, 2026 | Flexible audience updates and exclusions without a match threshold | The operations are active, while inclusion and bid multipliers retain their match requirement |
| June 16, 2026 | Conversion-optimized bidding | Live in the documented configuration and based on post-click conversions |
| June 11, 2026 | Expanded product, country, and device reporting | Data appears when the campaign, feed, and requested dimensions are eligible |
What is still rolling out or needs confirmation
Geographic expansion and self-service access should not be mistaken for uniform delivery. A market may be part of a rollout wave while a new account still faces review steps, different billing methods, or limited inventory. For a Romanian advertiser, the useful confirmation comes from the real account: Romania is selectable, currency and timezone are correct, billing is accepted, creative passes review, and the campaign starts reporting impressions.
Optimization that includes view-through conversions is not the current model
OpenAI has announced a conversion model expected to optimize using both click-through and view-through outcomes with impression-based billing. It should not be confused with the Maximize Results strategies documented today. In current reporting, view-through conversions are supplemental and do not change bidding, billing, or click-based CPA.
Until the new model has complete documentation and appears in an advertiser's account, a company should not forecast performance around assisted view conversions. It can preserve the data as a secondary signal, monitor attribution windows, and avoid adding results mechanically when another channel may claim the same conversion.
Uniform plugin and control availability
Even after a plugin or account control has been announced, its presence may depend on plan, role, region, organization, and Ads account status. People in the same workspace can have different permissions. The launch process should therefore remain workable without relying on one conversational command. Material changes must be confirmed on the campaign objects themselves and in reporting.
Total-budget pacing should be observed, not treated as a setting
Advertiser communications described more even pacing for campaigns with a total budget. The API documentation exposes daily and lifetime campaign budgets, but it does not present a separate advertiser-controlled pacing field. The prudent interpretation is that the platform can apply this delivery behavior and the team should observe it in spend data. It would be inaccurate to promise identical daily spending because eligible opportunities can vary from one day to another.
Five claims an advertiser should avoid
- “ChatGPT Ads is available, so our ad will appear immediately.” Delivery depends on review, schedule, budget, bidding, targeting, and inventory.
- “Every account has the same features.” Feeds, spending controls, and plugin access may depend on entitlement, billing, or rollout.
- “View-through conversions already optimize the campaign.” They are reported separately; the model expected to use them for optimization has been announced for a later stage.
- “If a capability exists in the API, it is available in every interface.” Technical documentation and a particular account's access are not synonymous.
- “An active campaign proves every component is eligible.” Creative, feed, product, conversion event, and billing can each carry a different state.
How to verify availability before committing budget
Instead of copying a list of features from an announcement, an advertiser needs a repeatable verification process. It can be completed before every creative asset is produced and before a material budget is reserved.
- Confirm account and role. Verify the organization, ad-account ID, connected user, permissions, currency, timezone, and billing type.
- Read the actual capabilities. Record the objectives, bidding strategies, targeting controls, feeds, and spending controls visible in the account. Include the verification date.
- Inspect dependent states. Review the account, campaign, ad group, ad, brand, feed, and products separately. A green status at a higher level does not override a rejection below it.
- Test measurement. Send a controlled event through Pixel and, when implemented, Conversions API. Verify deduplication ID, value, currency, and its appearance in reporting.
- Confirm geography and platform. Select Romania or the intended markets, then check app and web options. Avoid fragmenting the first test when expected volume is small.
- Run a bounded pilot. Use enough budget to create a signal while limiting downside. Track delivery, clicks, conversions, and the business result beyond the platform.
- Repeat after meaningful changes. A rollout, billing change, or new campaign type can alter availability. Internal documentation should always show its last verification date.
A simple decision matrix for Romanian businesses
| Situation | Recommended decision | Reason |
|---|---|---|
| The account has standard campaigns, local targeting, and working reporting | Begin a controlled pilot | The required core is available and measurable |
| The strategy depends on a feed, but feed access is absent | Do not build the full campaign around a carousel yet | Feed entitlement must be confirmed before production and forecasting |
| The plugin is not visible, but Ads Manager works | Continue through the interface or API if the team is prepared | The plugin is an operating method, not a prerequisite for the campaign itself |
| The plan depends on view-through optimization | Remove that assumption and use the current documented model | The new capability is announced, not confirmed as broadly active |
| Tracking cannot separate qualified leads from raw form submissions | Repair measurement before scaling | An active bidding feature cannot turn a weak signal into a sound business outcome |
Most companies do not need every available feature in the first test. A standard campaign, a clear landing page, one measurable outcome, and a controlled spending limit create a stronger foundation than a complex architecture built around capabilities that are still rolling out. Complexity should be added only after the data shows where it is necessary.
Keeping this map useful after launch
An availability map becomes stale quickly unless it has a date and an owner. The team should maintain a compact register containing the capability, observed status, account checked, official source, verification date, and campaign impact. When an update arrives, the register changes first; the sales presentation should not be updated from memory.
Three sources deserve continuous attention: product documentation, the update history, and the advertiser's own account data. Documentation explains rules, release records show what changed, and account data confirms what the company can actually use. None is sufficient alone. A feature can be documented but unavailable to the account, visible but unable to deliver, or delivering without creating commercial value.
For an assisted pilot, Web Hat's ChatGPT Ads campaign service can cover access verification, campaign setup, measurement, and result interpretation. Media budget remains separate, and the feature set is chosen after capabilities appear in the account and can be validated, not solely from product announcements.
Official sources checked
- OpenAI Ads, product overview and documentation entry point
- Advertiser API Overview, objects, access, and capabilities
- Account Management, account-level limits and controls
- Campaign Management, hierarchy and states
- Campaign Targeting, platforms and audiences
- Platform Targeting, app and web delivery options
- Custom Audiences, member updates, inclusion, and exclusion
- Product Feeds, access and product requirements
- Conversion-Optimized Campaigns, the active optimization model
- Bidding and Budgets, strategies and compatibility rules
- Reporting, conversions, attribution, and product data
Feature status was checked on September 16, 2026. OpenAI may change access, policies, interface controls, and technical specifications. Before making a budget decision, verify the current documentation and the configuration of the account that will run the campaign.
