Paused status is a state you set on a ChatGPT Ads campaign, ad group, or ad that stops it from serving while keeping its settings, history, and structure intact for later reactivation. Unlike archiving, which OpenAI treats as permanent, pausing is fully reversible, so you can switch the same campaign back to active without rebuilding it.
You reach for paused status when you need to stop spend fast, fix a problem, or hold delivery until a launch date, without losing the campaign you built. Treat it as an instant off switch and you can keep paying for ads that OpenAI allows to run for a short window after the change.
In the ChatGPT Ads hierarchy, paused status is the control state that halts delivery for a specific campaign, ad group, or ad while preserving every setting it holds. OpenAI's Ads API exposes three status values for a campaign: active, paused, and archived. Paused sits between the two, stopping delivery the way archived does, but staying reversible the way active allows.
You apply paused status in one of two places. In Ads Manager, you open the campaign and switch its status, then confirm the change on screen. Through the Ads API, you send the campaign to the pause endpoint or update its status field to paused. OpenAI's documentation states that paused campaigns will not deliver ads to customers, which is the behavior that defines the state.
Paused status is not the same as reaching an end date or exhausting a budget, both of which stop delivery on their own without any status change. It also differs from archiving, which OpenAI treats as irreversible. AirOps helps teams track how their brand performs across AI search and paid AI placements, so a pause becomes a deliberate decision backed by data.
See how AirOps brings intelligence and measurement to ChatGPT ads as a paid channel
Pausing is a status change you make on a single campaign, ad group, or ad, and it moves that object from active to paused in a few steps.
A paused status tells you the object is no longer entering auctions and that its configuration is preserved for later. It does not confirm every ad has stopped, so check delivery and spend reports for the day after you pause before assuming charges have ended.
Learn how AirOps syncs your ChatGPT ads account and manages campaigns in one place
Paused status is where budget control meets reaction time in ChatGPT Ads. How fast you pause, and when, maps directly to the dollars you spend and the moments your brand appears beside an AI answer. Marketers weighing whether to invest in ChatGPT ads need confidence that they can stop spend on demand, and paused status is that control.
ChatGPT Ads applies status at the campaign, ad group, and ad level, so pausing a campaign stops every ad beneath it, while pausing a single ad leaves its siblings serving and lets you act at exactly the scope where the problem lives.
Paused is a reversible state you can switch back to active at any time with the original settings intact, which is what separates it from archiving, a change OpenAI does not let you undo once it is done.
Delivery and charges can continue for a short window after you pause, so a pause records your intent to stop while actual serving can lag behind, which is why you confirm the result in reporting.
Avoid treating a pause as a permanent fix. A paused campaign still sits in your account spending nothing but earning nothing, so set a date to revisit it and decide whether to relaunch with changes or archive it for good.
AirOps: Syncs your ChatGPT Ads account structure and daily performance so you can see which campaigns to pause and measure the result against your AI search data.
ChatGPT Ads Manager: OpenAI's console where you switch a campaign, ad group, or ad to paused and confirm the change.
OpenAI Ads API: Lets you set status to paused programmatically or call the dedicated pause endpoint across many campaigns at once.
The core difference is reversibility: paused status can be switched back to active whenever you want, while archived status is a one-way move that OpenAI does not let you undo. Both stop an object from serving, so on the surface they look similar. The gap shows up the moment you want the campaign back. A paused campaign returns to active with its budget, targeting, and creative exactly as you left them, ready to serve again in seconds. An archived campaign is closed for good, and recovering its structure means rebuilding it from scratch or cloning it into a new object. Because of that, paused status is the right choice whenever there is any chance you will run the campaign again, and archiving is reserved for campaigns you are certain are finished. A simple rule: pause when you might come back, archive only when you never will.
You should not treat paused status as an instant off switch. OpenAI's advertising terms allow ads to keep delivering for a short window after you change a campaign, so delivery can tail off instead of stopping the second you click pause. In practice, assume a little spend may still land after your pause and that serving does not halt immediately. To know exactly what happened on your account, watch the delivery and spend reports for the day following the change and reconcile them against when you paused. If you need delivery to stop by a hard deadline, pause well before that deadline instead of at the last minute. Build that buffer into any brand-safety or budget plan, because the gap between your click and the final impression is the part you do not directly control.
Residual spend after a pause usually comes down to three things: the delivery window, the scope you paused at, and reporting timing. First, serving does not cut off the instant you pause, so impressions already in flight can still cost money. Second, if you paused one ad or ad group but left the parent campaign active, the rest of that campaign keeps spending as designed. Third, reporting can lag, so spend that occurred just before or during your pause may post to your account a little later and look like it happened after. The amount varies with how much volume your campaign was driving, how many objects you paused, and the time of day you acted. None of this means the pause failed. It means delivery and billing settle over a short period, which is why you confirm the outcome in reporting a day later instead of judging it in the first few minutes.
Not completely, and that is the honest answer. You directly control when you trigger the pause and at what scope, but you do not control the exact moment the last impression stops, because OpenAI winds delivery down on its own schedule after a change. What you can influence is the size of that gap. Pause earlier when a hard deadline matters, so the wind-down finishes before the date you care about. Pause at the narrowest scope that solves the problem, since stopping a single ad takes effect across less inventory than stopping a whole campaign. Keep campaigns you might need to stop quickly at a volume you can realistically monitor, because a smaller, well-watched campaign is easier to control than a sprawling one. You cannot make a pause instant, so plan for the lag and treat the pause as the start of the stop, with confirmation coming from your reports.
A good benchmark is simple: you should be able to pause a problem campaign within minutes of spotting it, while a response measured in hours is too slow. Because OpenAI keeps serving for a short period after a change, every minute you wait before pausing is added to the delivery you cannot stop, so speed is the metric that matters most. Strong teams set a clear threshold in advance, such as a cost-per-result ceiling, and empower more than one person to pause when that line is crossed, so a pause never waits on a single approver. They also know exactly where the controls are, which turns a pause into a reflex instead of a search. A second marker of good practice is how cleanly you reverse course: reactivating the right campaign with the correct settings, and confirming the status, should take under a minute. If pausing or resuming regularly takes you longer than that, the fix is usually clearer thresholds and a documented routine, and seldom more tooling.