OpenAI makes automated bidding the default in new ChatGPT Ads ad groups

OpenAI has made an automated bid strategy the default for new ad groups in ChatGPT Ads. The strategy is called Maximize results, it sets and adjusts bids on the advertiser's behalf, and it carries no cost target. PPC Land reported on the change, noting that OpenAI listed it first among six items in a product update email sent to advertisers in the ChatGPT Ads Manager Beta during the week beginning Monday 17 August 2026 in its coverage of the bidding default. That was the same week the pilot switched on in Mexico and Brazil, and the week before European inventory was scheduled to open. The help centre page behind the strategy sits in the Campaign setup collection of the ChatGPT Ads documentation and was stamped as updated in the second week of August 2026, which puts the technical documentation ahead of the note that told advertisers it existed.

The same email carried a second change. ChatGPT Ads now has a platform targeting control, a multi-select dropdown that restricts campaign delivery to iOS App, Android App or Web. The two changes belong to one story. OpenAI is moving ChatGPT Ads away from a manual product where the advertiser names a price and buys every surface, toward an automated one where the platform sets the bid and the advertiser gets a delivery switch it cannot yet fully measure.

What the Maximize results bid strategy does

According to OpenAI's documentation, Maximize results lets ChatGPT Ads set and adjust bids automatically, with the aim of returning as many results as possible from a given budget. It optimises toward the goal already selected at campaign level, which the documentation gives as clicks or conversions. The budget is the constraint on spend, and it is the only one.

Optimising for result volume is not new in paid search. Strategies that spend the budget and chase a count have existed for years across the established platforms. What is new here is where the strategy sits in the setup flow. For eligible new ad groups, Maximize results is preselected. An advertiser who wants something else changes it in the ad group's Bid strategy section.

OpenAI published a five-step setup path that makes the sequence explicit: create a campaign in Ads Manager Beta, continue to the ad group step, review Maximize results as the preselected option for eligible new ad groups, choose Manual: Max bid instead if a different strategy is wanted, then complete the ad group and ad setup and launch. Steps three and four are where control over the bid changes hands.

The default matters more than the strategy

Defaults decide behaviour at scale, and this one is presented in an unusual way. An advertiser who reads every field in a campaign builder will see the preselection and make a deliberate choice about it. An advertiser building a fourth ad group at the end of a working afternoon will accept whatever is already selected.

Nothing in the documentation describes a notification when the default applies, an account-level preference, or a way to set the account default to manual once and have it hold. The choice is presented per ad group, every time. For an agency running many accounts, that is a process problem rather than a settings problem. There is no switch to flip once and forget. There is a step in a build checklist that somebody has to perform on every ad group they create, and the cost of forgetting it is an ad group bidding without a ceiling.

What the strategy explicitly does not promise

The most consequential line in the documentation is a limitation. OpenAI states: "Maximize results prioritizes result volume and does not guarantee delivery against a specific CPA, CPC, ROAS, or other cost-efficiency target at this time." Advertisers who need a hard ceiling on what a click or an impression costs are directed to Manual: Max bid.

That sentence separates Maximize results from most of the automated bidding sold elsewhere in the market. Target-based strategies express a number the system is meant to hold. Maximize results expresses a count. Whatever cost per outcome falls out of the auction is the cost per outcome, and the advertiser finds out afterwards.

The qualifier "at this time" points toward a roadmap without committing to one. Neither the documentation nor the reporting said when, or whether, cost-efficiency targets will arrive alongside the volume strategy. Treat that as an open question rather than a promise.

Analysis, not sourced fact: volume maximisation is easiest to evaluate when an advertiser already knows what a reasonable cost looks like in the channel. Advertising inside ChatGPT began in the United States on 9 February 2026, and the self-serve Ads Manager opened to all United States businesses on 5 May 2026. A channel that young does not give most accounts a stable cost benchmark to judge the output against, which is precisely the situation in which a strategy with no cost target is hardest to supervise.

The second change: choosing which surface you buy

Platform targeting arrived as a multi-select dropdown with three supported values: iOS App, Android App and Web. It was documented in a help centre article revised within 24 hours of capture and listed in the same product update email as the bidding default.

Until this control appeared, buying a placement inside a ChatGPT conversation meant buying every surface the assistant runs on. Advertising began in the United States on 9 February 2026 covering mobile and desktop from the first day, and there was no way to buy one and not the other. The default remains unrestricted delivery, so a campaign created without touching the field behaves exactly as campaigns have behaved since February.

One of the three values is compound, and this is the detail most likely to catch a buyer out. OpenAI's documentation is explicit that Web covers both desktop web and mobile web. An advertiser who wants to reach only people on a laptop cannot express that through this control, because selecting Web admits a phone browser just as readily. The clean separations available are the two native applications versus everything else, and iOS versus Android inside the application layer.

There is also a naming inconsistency between the two documents. The help centre calls the field the Platforms multi-select dropdown. The product update email calls it Eligible platforms. Neither document resolves what the interface label actually reads, so an advertiser looking for the setting should expect either wording.

The reporting has not kept up with the buying

This is the most useful caution in the release. OpenAI states that Insights currently groups device reporting into Mobile and Desktop, with mobile web counted under Mobile, and that more detailed platform reporting is planned. No date accompanies that statement.

Three values go into the buying interface and two come back out of the reporting. An impression served in the iOS app and an impression served in the Android app land in the same Mobile bucket, so an advertiser cannot confirm from Insights how spend divided between the two applications. An impression served in a mobile browser, which the buying side classifies as Web, also lands in Mobile. Only a desktop browser impression reaches the Desktop row.

The practical result is that the new control can be set but its effect cannot be read back out of the platform's own reporting at the same resolution. The targeting boundary runs along the application axis. The reporting boundary runs along the device axis. They cross, and the advertiser is left holding a lever with no matching gauge until the promised reporting arrives.

Timeline of the changes

The table below lays out the documented dates from PPC Land's reporting and the OpenAI help centre pages it cites, against what changed on each of them.

What this means for Thai marketers

Start with the honest part. Thailand is not named in either source document as a live market for ChatGPT advertising. The markets identified in the reporting are the United States, the pilot countries the source lists, and the European expansion scheduled for the week after 17 August 2026. If you buy media in Bangkok, this is not a change you can act on inside a ChatGPT Ads account today, and anyone telling you otherwise is going beyond what the documentation says.

The transferable point is the pattern, and this is analysis rather than sourced fact. Automated bidding as a preselected default is the direction Google and Microsoft have already taken with their own campaign builders, and the operational lesson from those platforms holds here: the setting you never chose is still the setting you are running. Build reviews on Thai accounts are worth structuring around the bid strategy field specifically, because it is the field most likely to have been accepted rather than decided.

The second lesson concerns measurement discipline. A platform that offers a targeting control before it offers the matching report is asking advertisers to run changes they cannot verify. Teams working on GEO and AI SEO are already dealing with the same shape of problem on the organic side, where assistant answers can send qualified visitors without leaving a clean trail in conventional analytics. The habit that helps in both cases is the same: before changing a delivery setting, write down what number you expect to move, and check first that the platform actually reports it.

For Thai accounts spending on search today, the practical work sits in Google Ads, where automated bidding is mature, targets exist and the reporting supports them. Watching how ChatGPT Ads assembles the same control set is useful preparation rather than an immediate task.

What to check this week if you have access

1. Open the Bid strategy section on every ad group created since the second week of August 2026 and record which strategy is actually running. The preselection applies to eligible new ad groups, so recently built ad groups are where the default will have landed.
2. Decide, in writing, which campaigns are allowed to run without a bid ceiling. Manual: Max bid is the only documented route to a hard ceiling, so anything that needs one has to be set to it deliberately.
3. Add the bid strategy field to your ad group build checklist. There is no account-level preference to carry the decision for you.
4. Before you use platform targeting, confirm which conversion measurement you actually have in place for each surface, then check whether the report you plan to judge the change on exists. Insights returns Mobile and Desktop only.

What the sources did not say

• Whether existing ad groups are affected. The documentation describes the preselection for eligible new ad groups and says nothing about strategies changing on ad groups that already run.
• When cost-efficiency targets might arrive. The phrase used is "at this time" and no timeline was published.
• When the more detailed platform reporting will ship. OpenAI says it is planned, with no date.
• Whether platform selection can be edited after a campaign is live, which leaves open whether a surface split needs duplicate campaigns.
• What the interface label for the platform field actually reads, given the help centre and the email use different names.

Frequently asked questions

Is ChatGPT advertising live in Thailand?

The sources did not say that Thailand is a live market, and neither document names it. The reporting covers the United States launch on 9 February 2026, the pilot markets including Mexico and Brazil from 17 August 2026, and a European expansion scheduled for the following week. Treat Thai availability as unconfirmed until OpenAI documents it.

Do I have to do anything right now?

Only if you build ad groups in the ChatGPT Ads Manager Beta, in which case check the Bid strategy section on each new ad group. Everyone else has nothing to change today. The value of the story for Thai advertisers is the pattern it confirms about automated defaults, not an action inside an account.

Does Maximize results hit a target CPA or ROAS?

No. OpenAI states that it prioritises result volume and does not guarantee delivery against a specific CPA, CPC, ROAS or other cost-efficiency target at this time. Advertisers who need a hard bid ceiling are directed to Manual: Max bid instead.

Can I target desktop only with the new platform control?

No, because Web covers both desktop web and mobile web in OpenAI's documentation. The separations you can express are the two native applications against everything else, and iOS against Android.

Will my existing ad groups switch to the new strategy automatically?

The source did not say. The documented behaviour is preselection for eligible new ad groups, and nothing in the material describes a change applied to ad groups that are already running. If you manage a ChatGPT Ads account, verifying this yourself is faster than waiting for a clarification.

If you want a second pair of eyes on how automated bidding defaults are behaving across your search and AI channels, our team reviews account structure, bid strategies and measurement together rather than one at a time. Get in touch and we will start with what your accounts are actually set to.