OpenAI ChatGPT Ads Manager: Costs, Billing and What Changed in 2026

ChatGPT Ads Manager Is Live: What Changed on May 5, 2026Link to this section

Updated October 14, 2026.

ChatGPT Ads Manager is OpenAI’s self-serve advertising platform at ads.openai.com. It opened to all US businesses on May 5, 2026, with no minimum spend, CPC and CPM bidding, a Conversions API, and pixel-based measurement [1]. Ads run inside ChatGPT conversations for Free and Go users. They started in the US, Canada, Australia, and New Zealand [3], and by late September 2026 OpenAI was selling them in more than 60 countries [20].

If you read our original take in March, the world we predicted “in the next year” arrived in eight weeks. It has kept moving at that pace since.

What the self-serve launch actually unlockedLink to this section

The pilot was a $200,000 minimum filter. The self-serve beta is a different product. Five concrete changes matter:

• No minimum spend. The $200K floor became $50K in April and zero on May 5 [1] [2].
• CPC bidding. OpenAI’s recommended starting bid is $3 to $5 per click; lower bids were reported to struggle for delivery at launch [4].
• CPMs dropped. The flat $60 pilot CPM became a reported range of roughly $25 to $60 depending on category [6], though some advertisers still report CPMs above $60 [17].
• Conversions API and pixel. Post-click measurement now exists, closing part of the data blackout we flagged in March [1].
• Partner ecosystem. Dentsu, Omnicom, Publicis and WPP (agency) plus Adobe, Criteo, Kargo, Pacvue and StackAdapt (technology) are partners, which means your current stack likely has an integration path [1].

OpenAI is targeting $2.5B in ad revenue this year and $100B by 2030 [2]. Read that as a commitment to keep lowering the friction, not a one-off announcement. On August 31, 2026, OpenAI said the ads business had passed a $1 billion annualized run rate, less than 200 days after launch [9]. A run rate is one recent month multiplied by twelve, not booked revenue, so read it as a signal of direction [10].

What Changed Since May 2026Link to this section

OpenAI has shipped advertiser changes steadily since May. These are the changes that affect how you plan a test:

• June 2026: conversion bidding. Campaigns built on the Conversions objective bid toward conversions while still billing per click. You need conversion events flowing through the pixel or Conversions API first [7].
• June to August 2026: more markets. Ads expanded to the UK, Japan, South Korea, Brazil and Mexico [21].
• By July 2026: customer-list audiences. Upload a customer list to include or exclude people at campaign level, or to raise and lower bids [19].
• August 24, 2026: Europe. Ads went live in 31 European markets, selected on conversation topic, general location and device type, not on a personal profile [8].
• August 31, 2026: self-serve beyond the US. Self-serve access opened in Europe, India, the Middle East and North Africa [10].
• Early September 2026: Ads Manager inside ChatGPT. OpenAI’s own Ads Manager plugin lets you build, edit and review campaigns in plain language, with a preview and a confirmation step before any change applies [13].
• September 16, 2026: Sponsored Agents and CRM links. Sponsored Agents, an ad that opens a separate labeled chat with a brand’s agent, entered testing with select US advertisers [11]. HubSpot and Shopify became the first CRM and ecommerce integrations, and attribution windows became adjustable [12].
• Late September 2026: Southeast Asia and Taiwan. The expansion took coverage past 60 countries [20].
• October 5, 2026: image ads and measurement partners. OpenAI said it will test ads shown during image generation in the US later in October, named third-party measurement partners for click attribution, full-funnel measurement and geo experiments, and began brand-suitability pilots with DoubleVerify and Integral Ad Science [14] [16].

ChatGPT Ads Pricing in 2026: CPC vs CPMLink to this section

Three buying models, same auction. The choice is about what you measure, not what is cheaper.

The pilot pricing was $60 CPM with a $200,000 monthly minimum, which translated to a real-world spend of about $2.4M per quarter to find out whether the channel worked [5]. The self-serve beta lets you answer the same question for under $5,000.

Every price above is early and directional. OpenAI says it does not yet have performance benchmarks across advertisers, industries or campaign types, and it does not guarantee a CPA, CPC or ROAS [17]. Advertisers who have shared numbers report CPCs anywhere from about $3 to $22 [17]. Your own test is the only benchmark that counts.

How the Auction WorksLink to this section

ChatGPT Ads runs a relevance-weighted, second-price auction [4]. Two things follow from that.

First, your bid is not what you pay. As in Google’s auction, you pay just above the next-highest effective bid, so the discipline is to bid your real maximum willingness to pay, then let the auction discount you down to fair market price.

Second, relevance is a multiplier, not a tiebreaker. OpenAI scores how well your creative fits the conversational context, and a tightly written ad on a $3 bid can outrank a generic ad on a $7 bid. OpenAI’s recommended starting bid is $3 to $5 per click, and early reports had lower bids struggling to deliver [4]. New ad groups now default to a “Maximize results” strategy that sets bids automatically, so switch back to manual bidding if you want a hard ceiling [17].

Practically, this means the work on ChatGPT looks more like SEO craft than Meta media buying. You optimize for conversational fit (which intents you want to be adjacent to, what tone the ad strikes, what the offer is) before you optimize for bid.

How ChatGPT Ads Are Billed, and How to Pay for ThemLink to this section

Most guides stop at the bid. The bill is where a new channel usually goes wrong, so here is how OpenAI’s billing documentation describes it [18]:

• Postpay. You are charged after your ads run, not before.
• A saved card and a payment threshold. A self-serve account saves a card. Each time unpaid spend reaches the account’s payment threshold, that card is charged. The threshold may rise as you build a payment history.
• The threshold is not a budget. It is separate from your campaign budget and it is not a spend cap. It decides when you are charged, not how much you can spend.
• A temporary hold. Adding a card places a temporary authorization hold to verify it. It is not a charge.
• Invoicing. Eligible sales-managed advertisers can pay on invoice terms instead.
• Failed payments stop ads. If a charge fails, delivery can pause until it is fixed.

We found no prepaid balance option in that documentation as of October 2026.

Two practical consequences. Threshold billing means many small charges through the month, not one tidy invoice, so reconciling spend against what Ads Manager reported is real work. And a failed charge can pause delivery. If the same card also pays for Google and Meta, a limit reached on one platform can cause a decline on another.

That is why we give each ad account its own card. The loop we care about is simple: an ad charge becomes a deal, and the deal becomes cash. You can only follow that loop if you know exactly which charge paid for which campaign.

Flyweel Cards¹ are virtual Visa cards for US businesses, locked to advertising and marketing, each with its own cap. They work on ad platforms that accept Visa, so you can use one as the saved card on a ChatGPT Ads Manager account, the same way you would on Google Ads or Meta. That keeps the channel’s charges on a card that pays for nothing else.

A new channel also means paying for ads before you know what they return. Flyweel Capital funds ad spend with terms built around your sales cycle, for the weeks when the ads are working but the cash is not in yet.

What This Means for Your PipelineLink to this section

You can stop treating ChatGPT as a 2027 problem. Three operational implications:

Budget shape. The cost of a real test dropped from “fly a CFO to New York” to “a routine line item”. A reasonable diagnostic is 2 to 4 weeks of CPC spend at $5,000 to $15,000, instrumented to your CRM. If you cannot close the loop on that budget, the channel is not the problem.

Measurement. The Conversions API is the unlock. Pipe ChatGPT-attributed clicks into your CRM the same way you pipe Meta Ads or Google Ads, and judge them on cost-per-opportunity rather than cost-per-click. Platform dashboards still inflate; your CRM is the truth.

The data blackout is partial, not gone. OpenAI still does not share chat prompt content with advertisers. What changed is the post-click side: you now get the same kind of conversion signal that Meta’s CAPI gives you. You are still flying blind on pre-click intent, just no longer flying blind on outcomes.

Measurement is the weak point, and it is being worked on. On October 1, 2026, Digiday reported that one UK agency saw form fills arrive while Ads Manager still showed zero conversions two weeks later, and that many advertisers have not switched on conversion tracking at all because their legal teams will not accept the terms. Other agencies in the same report saw no discrepancies. Budgets are staying at test level as a result [15]. Four days later OpenAI named its third-party measurement partners [16]. Set up your own tracking before you spend, and do not rely on the platform to grade itself.

The honest framing: ChatGPT Ads is no longer an experimental brand play. It is the third paid channel sitting next to Search and Social, and like both of those it rewards operators who run it as a P&L line, not a vanity surface.

What We Got Right, and What Changed Faster Than ExpectedLink to this section

In March we made three predictions. The grades:

• “The $200K bar will drop as they build a self-serve portal.” Confirmed. It went $200K, then $50K in April, then zero on May 5 [1] [2].
• “They will need to offer CPC to win over performance marketers.” Confirmed. CPC bidding launched on day one of the self-serve beta [4].
• “Attribution will be a mess.” Half right. The Conversions API plus pixel measurement is real, and it landed two quarters earlier than we expected [1]. But the tools arriving is not the same as the numbers being trusted, and agencies are still reporting gaps [15]. There is still no prompt visibility on the targeting side either.

The lesson is one we apply to every new ad surface: when a category leader needs revenue, the timeline collapses. We expected a year of staged unlocks; OpenAI did it in eight weeks.

That pace is why we are bullish. In five months OpenAI went from four countries to more than 60, added conversion bidding and audiences, and answered agency complaints about measurement [15] with a partner list inside a week [16]. The limits are just as real. As one agency executive told Digiday, ChatGPT has one ad placement and far fewer daily prompts than Google has searches [16], and most advertisers are still spending test money [15]. Both things are true. A platform this early, shipping this fast, is worth a small, well-measured test now. It is not yet worth moving your Search budget.

Working From Inside ChatGPTLink to this section

You can now run parts of this from a ChatGPT conversation. OpenAI’s Ads Manager plugin builds and edits ChatGPT campaigns in plain language [13].

The Flyweel app in ChatGPT covers the other side: what you spent and what it earned. It reads your Google Ads, Meta Ads and TikTok Ads data, your CRM pipeline, and your Flyweel card charges. An owner or admin can freeze a card or set its budget from the same chat, after confirming. The same tools are open to Claude, Cursor and terminal agents through the Flyweel MCP. We call that headless finance for ads.

One limit to be clear about: Flyweel does not pull campaign data from ChatGPT Ads yet. A connector to OpenAI’s ads API is planned. Until then, Flyweel sees ChatGPT ad spend only as charges on a Flyweel card, if that is how you pay for it. There is no campaign-level reporting.

What To Watch NextLink to this section

A handful of questions are still open. These are the ones that will decide whether ChatGPT becomes a top-five paid channel or stays a top-twenty:

Premium tier ad rolloutLink to this section

Ads currently appear only to Free and Go users. If you sell to enterprise buyers who live on Plus, Pro, or Enterprise plans, your audience is structurally limited. Watch whether OpenAI extends ads (or a sponsored-link format) into paid tiers [3].

Global rollout beyond US, CA, AU, NZLink to this section

This one moved fastest. The four English-speaking launch markets became more than 60 countries by late September 2026 [20]. The open question is now regulatory. In Europe, ads are chosen from the current conversation, not from a personal profile [8]. In early October 2026, data protection authorities in Hamburg, Berlin and Lower Saxony warned users about OpenAI’s plans for personalized ads [22]. Expect European targeting to stay narrower than US targeting for some time.

Agency vs self-serve cost gapLink to this section

Dentsu and Omnicom are launch partners, and the typical pattern is that agency desks get inventory, beta features, and pricing that self-serve does not. Watch whether running through a partner buys you a measurable CPC or relevance advantage. If yes, hybrid the buy.

Prompt-aware targeting without sharing promptsLink to this section

OpenAI’s open puzzle is how to let advertisers target on intent without exposing chat content. Federated signals, encrypted topic embeddings, and “interest cohorts” generated server-side are all plausible. The first vendor to crack this defines the next iteration of the platform.

Integration with major buying stacksLink to this section

Adobe and Criteo onboard the first slug of programmatic plumbing, and HubSpot and Shopify followed in September 2026 [12]. Watch for hooks into Search Ads 360, Skai, and the major MMM vendors. Until then, expect to bolt ChatGPT into your stack manually.

Sponsored Agents pricingLink to this section

Sponsored Agents are in a US test with no public pricing or measurement detail [11]. If a click can open a conversation with your own agent, the landing page stops being the only place a lead gets qualified. How OpenAI charges for that will decide who can afford it.

Inventory beyond one placementLink to this section

One ad placement limits how much budget the platform can absorb. The image-generation ad test in October 2026 is the first sign of a second surface [14]. More placements mean more supply, and more supply is what moves budgets past test level.

Ready to stop guessing whether your ads pay back? See how Flyweel ties ad spend to deals and cash across Google, Meta and TikTok today, with a ChatGPT ads connector planned.

Frequently Asked QuestionsLink to this section

How much does it cost to advertise on ChatGPT in 2026?Link to this section

ChatGPT Ads Manager supports CPC, CPM and conversion bidding. OpenAI recommends a starting CPC bid of $3 to $5 per click [4], and advertisers have reported CPCs from about $3 to $22 [17]. Reported CPMs run from roughly $25 to above $60 [6]. OpenAI publishes no benchmarks, so treat every figure as early and directional. There is no published minimum spend in self-serve.

Is there a minimum spend for ChatGPT Ads Manager?Link to this section

Not in self-serve. The $200,000 monthly minimum that gated the March 2026 pilot dropped to $50,000 in April and was removed entirely when ChatGPT Ads Manager opened to all US businesses on May 5, 2026 [1] [2]. A meaningful test starts at the price of a small Google Ads campaign.

Does ChatGPT Ads support CPC bidding?Link to this section

Yes. CPC bidding launched alongside the self-serve Ads Manager on May 5, 2026 and runs in parallel with CPM [4]. Conversion bidding, still billed per click, followed in June 2026 [7]. The underlying relevance-weighted second-price auction means a strong ad on a smaller bid can win against a weak ad with a larger bid.

Can you track conversions from ChatGPT ads?Link to this section

Yes. The platform ships with a Conversions API and a pixel-based measurement option, so you can tie post-click events (signups, purchases, demos) back to the impression that drove them [1]. Some agencies have reported gaps between platform numbers and their own tracking, so run your own measurement alongside [15]. The user’s chat prompt content is still private; what you get is post-click signal, not pre-click intent.

Who sees ads in ChatGPT?Link to this section

Ads appear to Free and Go tier users. They started in the US, Canada, Australia, and New Zealand [3], and by late September 2026 were running in more than 60 countries [20]. Plus, Pro, Business, Enterprise and Edu users remain ad-free for now. If your buyers skew toward power users on paid plans, your reachable audience is narrower than the headline ChatGPT user count.

How is ChatGPT Ads different from Google Search Ads?Link to this section

Two material differences. You bid into a conversation, not a query, so intent is inferred from conversational adjacency rather than a typed keyword. And OpenAI does not share the prompt content with advertisers, so targeting is broader and post-click measurement carries more of the analytical weight. Practically, it sits between contextual display and intent-based search.

How do you pay for ChatGPT ads?Link to this section

ChatGPT ads are postpay. According to OpenAI’s billing documentation, a self-serve account saves a card, and that card is charged each time unpaid ad spend reaches the account’s payment threshold [18]. The threshold is separate from your campaign budget. Eligible sales-managed advertisers can be invoiced instead. A failed payment can pause delivery.

The Bottom LineLink to this section

The launch we wrote about in March was a press release. The launch on May 5 is a product. CPC bidding, no minimum spend, and a Conversions API turn ChatGPT from a closed brand experiment into a normal paid channel, one you can A/B against Search and Social on its own merits.

Advertising is still a timing game. You spend cash today to earn revenue later, and the channels that win are the ones where you can prove the round trip. The AdSpendOps clarity layer handles the reporting and reconciliation work across the channels it connects, when platform-reported numbers start drifting from bank-reported numbers. Advertising Finance (AdFi) is the broader system that also covers controls and capital. Run a small CPC test, instrument the Conversions API on day one, and let the CRM tell you whether the auction is paying you back.