ChatGPT Ads drops Max Bid for Target CPA on November 9. A CPG brand's plan for the switch

On November 5, 2026, OpenAI emailed ChatGPT advertisers about a change that starts on Monday, November 9. Search Engine Roundtable published the email, and OpenAI's help page on bidding now spells it out. New conversion campaigns get Target CPA in place of Max Bid. New click campaigns get Target CPC in place of Max CPC. The old controls retire for everyone on January 1, 2027.

This is a familiar move for anyone who has run Google or Meta ads: hand the auction to the system, give it an average cost to aim for, and stop setting bids by hand. The timing is the part that deserves care. The switch lands three weeks before Black Friday, on a platform where most CPG advertisers have only a few months of data, and where the early evidence on returns is not flattering. This playbook is the order we would do things in.

What changes, and when

Date	What happens
November 9, 2026	Target CPA replaces Max Bid for new conversion campaigns; Target CPC launches for new click campaigns. Existing campaigns can switch from this date.
November 18, 2026	OpenAI recommends API partners use the new Target CPA and Target CPC fields.
Through December 31, 2026	Legacy API fields still work, but using them to create a campaign selects the target strategy, and submitted values are read as average-cost targets.
January 1, 2027	Max Bid and Max CPC retire for existing campaigns and are deprecated in the API.

Source: OpenAI Help Center and Search Engine Roundtable.

The third option, formerly Maximize Results (Autobid), is now called Intelligent bidding. The Keywordless has a clear walk-through of the dates if you need to brief an agency. It pursues your conversion or click goal within the budget, with no target at all.

The detail that changes how you set a number

The most important sentence in OpenAI's documentation is the one about what a target is not. Target CPA and Target CPC are average-cost goals. They are not caps. Under Max CPC, no click could cost more than your bid. Under Target CPC, OpenAI says the system can bid "above or below your target in each auction" and that "actual average CPC may be higher or lower."

OpenAI gives a worked example worth copying into your planning doc. If a campaign spent $400 and recorded 10 attributed purchases, its actual CPA was $40. Setting a $20 target, OpenAI writes, "does not mean the system can deliver purchases at half the observed cost." A target below recent performance may simply restrict delivery and leave budget unspent.

The same page notes that the bidding goal and the billing event are separate. A conversion-optimized campaign is still billed on clicks or impressions, depending on what you chose when you created it, and OpenAI's conversion campaign guide says neither the objective nor the billing option can be changed in place. Moving from click billing to impression billing means cloning the campaign.

Why CPG brands should go slowly

An average-cost target works when the system has enough conversions to learn from. That is where packaged goods are at a disadvantage on any platform, and ChatGPT more than most.

An observational working paper from researchers at WU Vienna and Frankfurt School of Finance and Management, covered by Search Engine Land, compared paid ChatGPT visits with Google Search ad visits across sites in its sample.

Show the data

Paid channel	Conversion rate
Google Search ads	1.17%
ChatGPT ads	0.37%

Source: Search Engine Land, Nov 2, 2026, reporting a working paper by WU Vienna and Frankfurt School. Observational study of 1,898 websites, June 2025 to August 2026; paid comparisons used a smaller sample. Revenue attributed to the visit may miss later purchases.

Paid ChatGPT visits generated $0.89 per session against $2.45 for Google Search ads. Of 87 advertisers using both channels, 97% earned less revenue per ChatGPT visit. The researchers estimated ChatGPT ads would need to cost about $0.26 per click to match Google Search's return on ad spend. We covered the full study in what a paid ChatGPT visit is worth.

Measure	ChatGPT ads	Google Search ads
Revenue per session	$0.89	$2.45
Conversion rate	0.37%	1.17%

Source: Search Engine Land, November 2, 2026.

Two cautions. The study measures revenue attributed to the visit, which misses later purchases, and a CPG product bought at a grocery store a week later never shows up at all. And it is one study. But it points to the practical issue: at a conversion rate in that range, a small DTC budget produces a handful of purchases a week. That is thin data for any target-based system, and a target set from a handful of purchases can swing.

The plan

1. Before November 9: fix the signal

OpenAI's own checklist starts with conversion data, and so does ours. Confirm that the OpenAI Pixel, the Conversions API, or both, are sending the event you optimize for to the data source linked to the campaign. OpenAI is explicit that setup alone does not prove usable signals are arriving. Check the Event Quality Score and the diagnostics for missing matching information, delays and repeated events.

For a CPG brand, the question is which event. A purchase on your DTC site is the honest choice but may be rare. Conversion campaigns support one standard event per campaign, and custom events are not supported for the conversion objective. Do not optimize toward a softer event, such as add to cart, just to get volume, unless you are prepared to judge the campaign on that event too.

2. Pull your real numbers

For each campaign, take spend and attributed conversions over a consistent period, allow for reporting delay, and calculate actual CPA. Do the same for actual CPC on click campaigns. Write both down with the date range. That is your starting target, and it is also the baseline you will compare against after the switch. If you have fewer than a few dozen conversions in the period, treat the CPA as a rough range rather than a number.

3. Decide which campaigns move first

Existing Max Bid campaigns keep serving until January 1. You do not have to move them on November 9. We would leave a campaign that is performing acceptably alone through Cyber Monday and switch it in December, when a learning period costs less. New campaigns created from November 9 get the target options by default, so any holiday campaign you launch next week will be a target campaign whether you like it or not.

4. Set the target at, not below, recent actual cost

OpenAI's advice is to start from recent actual CPA and adjust as the campaign learns. Resist the urge to set an aspirational number. If your actual CPA was $40, a $30 target is a request to spend less, not an instruction to get cheaper purchases. If an audience bid multiplier is available, remember it scales the target: OpenAI's example is a $50 Target CPA with a 1.2x new-customer multiplier giving that segment a $60 target.

5. Keep a holdout

Leave one comparable campaign on the old setting through December, or keep a region or a product line out, so you can tell what changed because of the bidding and what changed because it was the holidays. Without one, every number in January will be a blend of season, budget and strategy.

6. Watch delivery, not just cost

The failure mode OpenAI warns about is under-delivery: a campaign that does not spend its budget because the target is too tight or the signal too weak. Check spend against budget daily for the first two weeks. A campaign spending 40% of budget at your target CPA is telling you the target is too low for the volume available.

7. Read the ad copy while you are in there

ChatGPT Ads turned on automatic creative changes by default in late October. If you sell products with regulated or sensitive claims, check what is being shown. We wrote about that risk in when ChatGPT rewrites your ads, and tip 35 has the short version.

What paid can and cannot do for you here

Paid placements in ChatGPT sit next to the unpaid answer, not inside it. OpenAI says product results in ChatGPT are selected independently and are not ads. So the switch to Target CPA changes how efficiently you buy attention in ChatGPT. It does not change whether the assistant recommends your product when nobody is paying. For most CPG brands we think that unpaid answer is still where most of the value is, and it is worth checking it before and after any paid push: run the same shopper questions, see whether your product is named, and see what is said about it. Some teams do this by hand; others use a tool such as SearchPanel from XENA Intelligence, which shows the actual answers for each product.

A one-page checklist

• Conversion event confirmed arriving at the linked data source, Event Quality Score reviewed.
• Actual CPA and CPC per campaign, with date ranges, written down.
• Decision recorded for each existing campaign: switch now, switch in December, or retire.
• Targets set at recent actual cost, multipliers noted.
• Holdout campaign or region chosen and left alone.
• Daily spend-against-budget check for two weeks after any switch.
• Ad copy reviewed for automatic changes.
• API integrations updated to the new target fields before January 1, 2027.

None of this is exciting, and none of it needs to be. The switch is mostly a test of whether your conversion data is good enough for the system to use. If it is not, fix that first. The bidding strategy comes second.