OpenAI is converting the fixed daily spending limits in ChatGPT Ads Manager into seven-day averages, a change applied to running campaigns without advertiser action and without a published single-day overspend ceiling. The switch was flagged in a weekly product update email to beta advertisers, reported on July 24, 2026, as arriving the following week.

The note listed eight changes. Seven of them add something: conversion-optimized bidding, geographic exclusions, mobile measurement integrations, a bulk API, hashed-data conversion matching, automatic intraday pacing, and a redesigned product card. One of them removes something. The daily budget, introduced nine weeks earlier as a hard per-calendar-day cap, stops being a cap.

"This week's product updates focus on helping advertisers optimize toward conversion goals, improve conversion measurement, pace budgets more effectively, and scale campaign operations," the company told advertisers, according to the email.

Budget pacing sits third in that list. It is arguably first in consequence.

What the daily budget change actually does

According to OpenAI, the mechanism is a redefinition rather than a new control. "Daily budgets will change from fixed daily limits to average daily budgets: the average you want to spend per day over seven days," the email states. It adds two reassurances: "No action is required, and your total budget is not changing." Spend, the note continues, "may vary by day while staying within daily and seven-day limits."

Three properties of that description carry weight for anyone running spend through the platform.

The first is the absence of an opt-in. Advertisers who selected a daily budget in the Ads Manager selected a per-day ceiling. That selection now maps to a different behaviour, and the email describes no mechanism for declining the conversion or reverting to the previous arrangement. The change reaches campaigns already in flight.

The second is the window length. Seven days is shorter than the monthly reconciliation period used by the largest incumbent platforms, which limits how far a campaign can drift from its nominal daily figure before the system corrects. A short window constrains cumulative divergence. It does not constrain any single day.

The third is what the phrase "daily and seven-day limits" leaves undefined. The seven-day limit is calculable: seven times the average daily budget. The daily limit is asserted but never quantified. Nothing in the email states how far above the average daily figure a single day's spend can run.

Intraday pacing arrives alongside

A second budget item in the same note addresses distribution within the day rather than across the week. According to OpenAI, "Daily budgets now pace automatically throughout the day, helping distribute spend more evenly."

Automatic pacing partially answers a problem the platform created for itself. When ChatGPT Ads Manager first gained daily budgets on May 22, 2026, the stated rationale was preventing a campaign from exhausting its funds in the opening hours of a cycle. That is a front-loading problem, and even pacing is the standard remedy. The May feature applied only to new campaigns; existing ones could not be switched retroactively.

Pacing and averaging pull in different directions. Intraday pacing smooths delivery within a twenty-four hour block. Seven-day averaging permits deliberate concentration of spend on days the system judges more productive. Running both means the platform distributes evenly inside a day whose total it has itself chosen to raise or lower.

The ceiling that has not been published

Comparison with established platforms makes the gap legible.

Google Ads has operated average daily budgets for years, and the arithmetic governing them is public. The system may spend up to twice the average daily budget on a single high-traffic day, with monthly billing capped at 30.4 times the average daily figure. That multiplier of 30.4 represents the mean number of days in a month, applied uniformly rather than recalculated for February or March. The two-times daily ceiling was restated in Google's own advertiser communications ahead of a March 1, 2026 pacing change, where the company committed that billing would never exceed twice the average daily budget in a single day.

Whatever else can be said about that architecture, an advertiser can compute the worst case. Multiply by two for a day. Multiply by 30.4 for a month. The numbers are contractual.

OpenAI's note supplies the seven-day multiplier implicitly and the daily multiplier not at all. An advertiser holding a $500 average daily budget can derive a $3,500 weekly ceiling. The maximum exposure on any individual Tuesday remains unstated.

The direction of travel is also worth noting because it runs against the recent movement at the largest search platform. Google spent the first half of 2026 building hard ceilings rather than dismantling them. Campaign total budgets, which fix a spending limit across a window of three to 90 days with no daily limit but an unbreachable total, expanded to Search, Performance Max, and Shopping campaigns in January 2026 after sustained advertiser requests for absolute spend control during promotional periods. Google reported that advertisers using them cut manual budget adjustments by 66% on average, a figure disclosed alongside a separate demand-led pacing feature in May 2026.

One platform added a fixed ceiling because buyers asked for one. The other converted its fixed ceiling into an average.

The seven other items

The budget change travelled with a set of additions that broaden what the platform can do rather than alter how it spends.

Conversion-optimized campaigns are now created by choosing the Conversions objective during setup. According to OpenAI, these produce conversion-optimized cost-per-click campaigns, abbreviated oCPC, which "optimize delivery toward clicks that are more likely to drive your conversion event, while billing remains based on valid clicks." Billing stays tied to clicks while delivery is weighted by predicted conversion probability, which places the forecasting risk on the platform rather than the buyer. The objective itself first rolled out on June 5, 2026 to accounts with a conversion event flowing through the pixel or Conversions API by June 1.

Geo exclusion permits specific locations to be removed from targeting, inverting the inclusion controls that arrived in May at state, designated market area, and ZIP code level. According to the email, exclusions align delivery "with your audience and business needs."

Mobile measurement partner integrations with AppsFlyer and Adjust were described as recently launched, covering app installs and in-app events. PPC Land examined those integrations and their seven supported markets in detail on July 26, including the deduplication of server-side and pixel events against a shared event identifier.

Automatic advanced matching, abbreviated AAM, improves website conversion measurement by matching additional conversions using hashed customer information. Unlike the budget change, it requires deliberate activation through Tools, then Conversions, then Data Source, followed by selecting the option to enable it for all web data sources. Hashed identifier handling on the platform already carries a floor: custom audiences require a minimum of 25,000 matched usersbefore they can be applied to a campaign.

Bulk API access allows campaigns, ad groups, and ads to be created and updated asynchronously through the Ads API. Bulk operations are the practical precondition for managing large account structures, and their absence has constrained how much budget agencies could reasonably administer inside the platform.

A refreshed product feed card format is beginning to roll out with price and star ratings on the creative. The sample in the email shows a cordless vacuum listed at $249.99 with a 4.6-star rating and a merchant name beneath it. Feed support moved into the Ads Manager in early June after the platform's instant checkout experience was discontinued.

Nine weeks from cap to average

The compressed timeline is what makes the budget item unusual.

ChatGPT advertising launched on February 9, 2026 for logged-in adults on the Free and Go tiers in the United States, priced at a $60 CPM with a minimum commitment between $200,000 and $250,000. Daily budget controls were listed as forthcoming when the Ads Manager first appeared in April with the threshold cut to $50,000. They were still absent on May 5, when OpenAI opened the self-serve beta to all United States businesses, removed the minimum entirely, and introduced cost-per-click bidding at a recommended $3 to $5 per click.

Daily budgets shipped on May 22. Sixty-three days later, they were redefined.

Feature velocity across that period has been high by any measure. Custom audience targeting surfaced in a gated rollout on May 14. LiveRamp connected its CAPI Hub on June 10. Advertising activated in the United Kingdom on June 6, then in Japan and South Korea on June 22 alongside self-serve access for British advertisers. Building a budgeting system in public produces exactly this pattern: a control ships, meets live auction conditions, and is reworked.

The reworking is not itself remarkable. The absence of a stated numerical bound on the new behaviour is.

Why this matters for the marketing community

Spend predictability is the practical question, and it is sharpened by what independent measurement has recorded about the platform's auction.

Adthena counted 7,378 distinct advertisers inside ChatGPT during the week of July 13 to 20, 2026, with the United States accounting for 60.1% of them and placements reported to have grown 97 times between April and June. That growth arrives on top of pricing that has already moved substantially: CPMs fell from $60 at launch to as low as $25 by mid-April as inventory outpaced demand.

Auction volatility followed the expansion. When the United Kingdom auction opened, brands encountered cost-per-click swings ranging from 39% to 278% with no visibility into which competitors had entered against them, a pattern documented in Adthena's partnership announcement with Dentsu. ChatGPT's native reporting returns an advertiser's own numbers and nothing about anyone else's.

Averaged budgets and volatile clearing prices interact. A system permitted to spend above the nominal daily figure will do so on days it identifies as high-opportunity, and in an auction where unit costs can move by triple-digit percentages without explanation, high-opportunity and high-cost are difficult to separate from the buyer's side. The weekly total remains bounded. The distribution inside that week does not, in any published sense.

For finance teams reconciling spend, the seven-day window is a meaningfully different accounting object than a calendar day. Media plans built on daily flighting, retail promotions tied to specific dates, and campaigns with hard cutoffs around events all assume a stable relationship between the number entered in the interface and the money leaving the account on a given date. That relationship has been loosened.

There is a counterweight worth registering. Averaging is the dominant convention across mature platforms for a defensible reason: rigid daily caps waste opportunity on high-demand days and leave budget unspent on quiet ones. Metaand Google both settled on averaged pacing after extensive operation at scale, and OpenAI adopting the same convention makes ChatGPT budgets behave more like the rest of a buyer's portfolio rather than less. The seven-day reconciliation period is also tighter than the monthly windows used elsewhere, which reduces how far spend can wander before it is trued up.

What separates the two cases is disclosure rather than design. Google's version comes with published multipliers that make the worst case computable before a campaign starts. The version described in OpenAI's email comes with a stated total, a stated averaging period, and an unstated daily bound.

The changeover was scheduled for the week beginning July 27, 2026. OpenAI has published no separate confirmation that the conversion completed, and no numerical daily ceiling has appeared in the material circulated to advertisers.

Timeline

Summary

Who: OpenAI, operator of the ChatGPT Ads platform, communicating to advertisers enrolled in the ChatGPT Ads Manager Beta. The budget change affects every advertiser running a daily budget on the platform, across the seven markets where ChatGPT advertising currently operates.

What: Fixed daily budget limits are being converted into average daily budgets, defined as the average spend per day across a seven-day period. Total budgets are unchanged and no advertiser action is required, but daily spend will vary while remaining within unspecified daily limits and a calculable seven-day limit. Daily budgets additionally began pacing automatically through the day. The same product note listed conversion-optimized cost-per-click campaigns, geographic exclusion targeting, AppsFlyer and Adjust integrations, automatic advanced matching for hashed conversion data, a Bulk API for asynchronous campaign operations, and a refreshed product card format carrying price and star ratings.

When: The weekly product update email was reported on July 24, 2026, describing the budget conversion as arriving the following week, which places the changeover in the week beginning July 27, 2026. Daily budgets were originally introduced on May 22, 2026, sixty-three days earlier.

Where: Inside ChatGPT Ads Manager at ads.openai.com, with the Bulk API operating through the Ads API. Automatic advanced matching is configured through Tools, then Conversions, then Data Source.

Why: Averaged pacing lets a delivery system concentrate spend on days it judges more productive rather than enforcing a uniform daily figure, a convention already standard at Google and Meta. The distinguishing feature of OpenAI's implementation is disclosure: Google publishes a two-times daily and 30.4-times monthly bound on its averaged budgets, while OpenAI's note states a seven-day averaging period and a total budget without quantifying how far a single day's spend can exceed the average.