ChatGPT Ads Is Retiring Max Bid and Max CPC on 1 January 2027 (What to Do Before Then)

ChatGPT Ads is replacing its two manual bid caps with cost targets. From Monday 9 November 2026, a new conversion campaign in Ads Manager uses Target CPA instead of Max Bid, and a new click campaign uses Target CPC instead of Max CPC. On 1 January 2027 the two old options stop existing, and any campaign still on them is moved across by OpenAI.

The practical difference sits in one phrase from OpenAI's own email to advertisers. Where Max CPC capped every single bid, Target CPC "lets our system bid above or below your target in each auction." The number you enter is no longer a ceiling. It is an average the system works toward, and OpenAI adds that a target set too low, or a conversion signal the system cannot read, can leave budget unspent.

For businesses in Indonesia, which gained self-serve access to ChatGPT Ads on 23 September 2026, this is the first structural change to the platform since launch. The sections below cover what OpenAI announced, why a target behaves differently from a cap, what to check before switching, and where paid placement sits next to the organic answer.

What OpenAI announced on 5 November

OpenAI told advertisers by email on 5 November that ChatGPT Ads is moving to cost targets, and its bidding strategies page confirms the dates. The email lists four changes.

• Target CPA launches for conversion campaigns and replaces Max Bid for new conversion campaigns in Ads Manager. You set the average amount you would like to pay for a purchase or a lead, and the system adjusts bids and delivery toward that figure.
• Target CPC launches for click campaigns and replaces Max CPC for new click campaigns. The system aims for as many clicks as the budget allows while working toward your average cost per click.
• Max Bid and Max CPC retire on 1 January 2027. Existing conversion campaigns still on Max Bid move to Target CPA, and existing click campaigns still on Max CPC move to Target CPC.
• Maximize results, the automated option also called autobid, continues under a new name. OpenAI now calls it intelligent bidding, and the email describes no change to how it works.

Existing campaigns keep serving on their current setting until the deadline. Nothing breaks on 9 November. The window between 9 November and 31 December is the period in which you choose each campaign's target yourself. After that OpenAI chooses for you, and the email does not say what value it will assign to a campaign it converts automatically.

API partners have a separate timeline. From 18 November the legacy Max Bid and Max CPC fields still exist, but using them to create a campaign or change an ad group's bid type selects the matching target instead, and the submitted value is read as an average-cost target rather than a maximum bid. Anyone running ChatGPT Ads through a tool or a partner platform should ask that partner what their integration does with existing values before 18 November.

Why a cost target behaves differently from a bid cap

A bid cap is a promise the platform makes to you about any single auction. Under Max CPC, a figure of 60,000 rupiah meant no click could ever be bought above that. Spend per click could land lower, but never higher, and that guarantee is what many advertisers used to protect a strict cost ceiling agreed with a finance team or a client.

A cost target is a goal the platform works toward across many auctions. OpenAI's help centre is explicit that Target CPA and Target CPC are average-cost goals and "not a fixed price per conversion or click." Some clicks will be bought above the target and some below, and the realised average "may be higher or lower" than what you entered. The system weighs the likelihood of a click or conversion against the price of the auction, then decides the bid on your behalf. Under the old setting your number was a wall. Under the new one it is a steering input.

The second difference is what happens when the target is too ambitious. OpenAI says the system adjusts "bids and delivery," and the word delivery matters. A target that the available auctions cannot support does not produce cheaper results. It produces fewer impressions, because the system declines auctions it judges unlikely to hit the average. The help centre gives the example directly. If a campaign spent 400 dollars for ten purchases, the actual CPA was 40 dollars, and "setting a 20 dollar target does not mean the system can deliver purchases at half the observed cost." The campaign will simply serve less and leave budget on the table.

What this means for advertisers in Indonesia

ChatGPT Ads opened self-serve buying to businesses in Indonesia, Malaysia, the Philippines, Singapore, Thailand, Vietnam and Taiwan on 23 September 2026, and on 21 October OpenAI extended the platform to 89 countries, saying more than 100,000 advertisers had run campaigns since the US test began in February. Any Indonesian account that set up a click campaign on Max CPC or a conversion campaign on Max Bid in the past six weeks is inside this change, and the email gives no market exceptions.

Who sees the ads is worth restating, because it shapes what a cost target can reach. Ads in ChatGPT show only to people on the Free and Go plans, and never to accounts OpenAI identifies as under 18. Plus, Pro, Business, Enterprise and Edu accounts stay ad free. In a market like Indonesia, where the free tier is where most consumer usage sits, that still leaves a large audience, but it is a consumer audience first. A B2B advertiser hoping to reach decision makers inside companies on Business plans will not find them here.

The auction itself is a relevance-weighted, second-price auction, which OpenAI describes in its ChatGPT Ads basics page. Relevance to the conversation, the landing page, the headline and the context hints an advertiser supplies all feed the score. A lower target with a highly relevant ad can still win. The ad appears below the ChatGPT response, carries the advertiser name and a label, and OpenAI says it does not influence the answer the person receives. That separation between the answer and the ad is where the rest of this piece ends up.

Checking conversion signals before you switch

Target CPA can only steer toward a cost per conversion if the campaign receives the conversion it is meant to optimise for. OpenAI's wording names three things that have to line up. The conversion event selected in the campaign, the data source linked to that campaign, and the OpenAI Pixel or Conversions API actually sending that event to that data source. A mismatch between any two of them leaves the campaign optimising toward an event that never arrives, and the help centre warns that "setup alone does not establish that usable conversion signals are being received."

The tool for finding gaps is the Event Quality Score, or EQS, shown in Ads Manager. It describes how useful the received event information is for measurement, flags missing matching information, delays and repeated events, and the Conversions page has been redesigned to show event volume, trends and diagnostic warnings together. OpenAI gives no threshold a campaign must reach before Target CPA works well. Any warning showing now should be resolved before the switch. A high score also does not mean the campaign has enough volume to learn from. Volume and quality are separate problems.

Weak signals have a direct cost under the new model. OpenAI's own answer to "will my campaign spend its full budget?" lists three reasons it may not. Targets that are too restrictive, limited eligible opportunities, and conversion signals that are weak or missing. Two of the three are inside the advertiser's control, and the signal check is the one with the most work in it. This is the same discipline the measurement side of a search program demands anywhere else, and a conversion event that fires on the wrong page in Google Ads fails in exactly the same way here.

Setting a realistic target from your actual numbers

OpenAI's three recommendations in the email end with the one that decides whether the transition works. Use recent actual CPA or CPC to choose the target, over a consistent reporting period, allowing for conversion-reporting delay. The help centre adds a second check, which is to compare that recent performance with what the outcome is worth to the business, because a target set below recent performance "may restrict delivery."

The arithmetic is simple and the temptation is the problem. A click campaign that ran on a Max CPC ceiling of Rp 60,000 will have a recorded average CPC somewhere below that, say Rp 38,000. The ceiling was never what the campaign paid on average. Carrying it across as a Target CPC would hand the system a looser goal than the campaign already achieves. The recorded average is the honest starting point. The same logic holds for conversion campaigns. If Rp 12 million of spend produced 40 leads, the actual CPA was Rp 300,000, and that figure, adjusted for what a lead is worth to you, is the target. Entering Rp 150,000 because that is what you would prefer to pay does not make the auctions cheaper, and the only effect is that the campaign serves less.

Audience bid multipliers, where available, stack on top of the base target rather than replacing it. A Target CPA of Rp 300,000 with a 1.2 times new-customer multiplier gives that segment a Rp 360,000 target, still as an average goal rather than a guaranteed price. Once a campaign is on the new strategy, OpenAI's advice is to review spend, results and actual average cost together while it learns, and to adjust the target in steps rather than resetting it to a wish. The bidding strategies page describes this as a learning process rather than a setting you fix once.

Deciding which campaigns to move and when

Between 9 November and 31 December every campaign still on Max Bid or Max CPC is yours to move or to leave. Moving it yourself, with a target drawn from its own cost history, keeps the decision with you. Leaving it for the 1 January switch hands the decision to OpenAI, whose email did not describe what target value it will apply. That silence is the strongest argument for moving campaigns early, and OpenAI's first recommendation is to "review your campaigns today."

A sensible order starts with a list. Every ChatGPT Ads campaign, with its current strategy noted as Max Bid, Max CPC or intelligent bidding, and its actual CPA or CPC over a recent consistent period. Conversion campaigns go through the signal check before anything else, because a Target CPA set against an event that does not arrive is the failure case OpenAI warns about. Click campaigns are simpler to move, since Target CPC needs only a realistic average rather than a working conversion feed. If a client relied on Max CPC as a hard per-click limit, that conversation needs to happen before the switch, because under Target CPC individual clicks can and will cost more than the number entered.

Several things the email did not cover are worth holding as open questions rather than assumptions. It says nothing about impression-based buying. Whether reach campaigns on a maximum CPM are affected is left open. It does not address the conversion-optimised CPC beta for product feed campaigns, and it gives no minimum conversion volume, learning period or ramp-up time for Target CPA. OpenAI's reporting in Ads Manager currently covers impressions, clicks, spend, click-through rate, average CPC, average CPM and conversions. That is enough data to watch a transition. OpenAI has not published the thresholds that make one succeed.

Where paid placement sits next to the AI answer itself

The ad appears below the answer, and OpenAI says the ad does not change the answer. That sentence describes the whole structure of visibility inside ChatGPT. There is the response, which cites and recommends whatever the model and its retrieval judge most relevant, and there is a labelled unit underneath it that a business can buy. A cost target steers the second. Nothing in Ads Manager touches the first.

The two are not alternatives, because they reach the same person at different moments. A business can hold an efficient Target CPA and still be absent from the recommendation the person just read, and for a buyer who already trusts the answer, the brand named inside it has a head start over the one placed beneath it. The work that puts a brand inside the answer is different work. It runs on the entity signals, third-party mentions and citable pages that AI systems retrieve from, which we describe in our AI search service and in the guide on how brands get cited in ChatGPT, Gemini and Perplexity. The Topic Ownership Strategy is how we decide which questions a brand should be the answer to before anyone buys a placement under them.

Paid placement in ChatGPT is still young, the targets are new, and the published guidance from OpenAI is thinner than what Google Ads advertisers are used to. What has not changed is the measurement discipline underneath it. A conversion event that reaches the platform, a target drawn from real cost history, and a report that ties spend to an outcome the business recognises. That is the same standard we hold our paid search work to, and the same standard we apply when we measure whether a brand shows up in the answer at all. If you run ChatGPT Ads in Indonesia and want a second pair of eyes on your conversion setup and targets before 1 January, or want to know whether your brand appears in the answers above your ads, talk to us.