OpenAI ad service reportedly bills paused campaigns for up to one business day after pausing

OpenAI’s ChatGPT Ads service includes a provision in its advertising terms that allows the platform to continue serving ads, and billing for them, for up to one business day after a customer pauses or cancels a campaign. This policy is not unique to OpenAI; similar terms exist across major ad platforms. However, the practical effect is that advertisers may face charges for impressions or clicks they did not intend to pay for, creating a gap between user action and system response.

The issue surfaced when a customer, Excel4Business, observed ads continuing to run after pausing campaigns, with one ad running nearly 94 minutes post-pause. OpenAI initially acknowledged this as a defect, attributing it to a delay in refreshing ad-level status after a campaign was marked as paused. The company later reversed its stance, citing its terms of service, which explicitly state that advertisers remain liable for ads delivered during the 24-hour window following a pause or cancellation.

For engineers and operations teams, this introduces a layer of complexity in ad spend management. Automated budgeting or cost-tracking systems must account for potential billing lags, particularly when reconciling daily spend against campaign settings. The discrepancy between the user interface (where a campaign appears paused) and the backend serving system (which may still deliver ads) highlights a common challenge in distributed ad platforms: state synchronization across microservices.

While OpenAI’s policy is framed as a standard practice, it contrasts with the real-time control advertisers often expect, especially in platforms like Google AdWords, where pauses typically take effect immediately. The difference may reflect OpenAI’s relative immaturity in the ad tech space, where latency in ad delivery systems is still being optimized. For teams integrating OpenAI’s ad API, this means designing for eventual consistency rather than immediate state reflection.

The broader implication is that ad platforms, regardless of vendor, may prioritize system reliability or scalability over real-time precision. Engineers should treat ad platform APIs as eventually consistent systems, where actions like pausing a campaign may not propagate instantly. This requires building in tolerance for delayed state changes in monitoring, alerting, and financial reconciliation pipelines.