OpenAI's advertising platform continues to serve — and bill for — campaigns after an advertiser pauses them, in some cases for up to a full business day. The behaviour is written into OpenAI's own contract, as reported on 13 August.
The clause
Section 11.1 of OpenAI's Advertising Terms states that ads "may continue running for up to one business day after a campaign is canceled or changed, and advertisers remain responsible for ads delivered during that period." OpenAI's position is that a pause counts as a change, not a cancellation — so the clause applies to it.
What one advertiser measured
UK agency Excel4Business, testing ChatGPT Ads in the US and Canada, documented delivery continuing about 94 minutes past a pause. It disputed £60.72 of charges on the first occurrence and a further £6.47 on a recurrence in early August. OpenAI first acknowledged the behaviour as a technical defect, then declined a full refund on the basis of the terms.
Why the framing matters
Calling this overcharging would be wrong. It is contractually disclosed behaviour that an advertiser agreed to, and the interesting question is not whether OpenAI breached its terms but why the terms permit a full business day of uncontrollable spend in a system that is otherwise real-time.
What nobody knows
There is no data on how often OpenAI actually uses the full one-business-day window, or how many advertisers are affected. One agency's £60.72 is an anecdote, not an aggregate, and OpenAI has announced no change to the clause.
The context around it
ChatGPT reported 900m weekly users and 50m paid subscribers as of February 2026. An ad business at that scale is being built on terms that shift a familiar piece of platform risk — spend after the stop button — onto the buyer.