First Campaign Results from the ChatGPT Ads Manager

How the ChatGPT Ads Manager works

OpenAI opened its ChatGPT Ads Manager to a limited pilot group in early April 2026, giving roughly 600 advertisers their first self-serve dashboard for buying ad placements inside ChatGPT conversations. [1] The minimum spend dropped from a managed-service floor of $200,000-$250,000 to $50,000, a signal that OpenAI wants volume, not just marquee logos, ahead of its planned 2026 IPO. [2] For anyone running paid media budgets, the practical question is whether this platform can produce performance data worth acting on, or whether it remains an expensive brand-awareness experiment with no real attribution story.

Structurally, the Ads Manager follows a familiar Campaigns > Ad Groups > Ads hierarchy. Each ad unit consists of a headline, description, and optional static image (no video support yet). Ads appear as sponsored cards below relevant ChatGPT responses, but only for Free and Go tier users in the US, with expansion underway to Canada, Australia, and New Zealand. [1] Paid ChatGPT subscribers, users under 18, and conversations touching sensitive topics like politics or health are excluded from ad delivery entirely. That exclusion logic matters: it means roughly 85% of ChatGPT’s user base is eligible, but fewer than 20% see ads on any given day. [3]

Before the Ads Manager existed, pilot advertisers received weekly CSV exports with aggregated impression and click data. Real-time reporting inside the dashboard is the single biggest functional upgrade, even though the metrics it surfaces remain limited to impressions and clicks. [4] There is no pixel-based attribution, no conversion tracking integration, and no connection to third-party analytics platforms. If you are used to the feedback loops in Google Ads or Meta, this will feel like flying with half your instruments taped over.

Campaign setup and targeting workflow

OpenAI’s targeting model breaks sharply from the keyword-and-audience paradigm that most paid media buyers have internalized over the past two decades. Instead of bidding on keywords or selecting demographic segments, advertisers write “context hints,” which are plain-language descriptions of the types of queries where their ads should appear. [1] You might enter something like “users asking about project management software for remote teams” rather than bidding on [project management tool] as a phrase match keyword. OpenAI’s model then decides, based on conversational context, whether your ad is relevant to a given exchange.

Geographic targeting exists, but only at the country level. There are no demographic filters, no behavioral segments, no retargeting lists, and no lookalike audiences. For advertisers accustomed to layering intent signals with audience data, this is a significant constraint. It means the entire targeting burden falls on how well OpenAI’s model interprets your context hints and matches them to live conversations, a process that is essentially opaque to the advertiser.

Campaign setup itself is straightforward. You create a campaign, define ad groups with context hints, set a daily or lifetime budget, choose start and end dates, and upload your creative assets. The only available objective at launch is “reach,” priced on a CPM basis. Click and conversion objectives are listed in the interface but grayed out as “coming soon.” [1] That said, some pilot users have reported access to CPC bidding at $3-$5 per click, suggesting OpenAI is testing performance-based pricing in parallel with the CPM model. [5]

Onboarding remains invite-only. Direct access requires the $50,000 minimum commitment, but partners like Criteo (which brought 17,000 commerce advertisers into the ecosystem on March 2, 2026) offer a programmatic path that may lower the operational barrier for mid-market brands. [2]

Initial CPC and conversion rate benchmarks

I want to be direct about the data situation: there are no verified public benchmarks for CTR, CPA, or ROAS from the ChatGPT Ads Manager pilot. What exists is a patchwork of reported figures, some from OpenAI’s partners, some from third-party estimates, and none independently audited.

Here is what we can piece together:

A few things stand out. The $3-$5 CPC range is high compared to many Google Search verticals but competitive in categories like fintech, insurance, or enterprise SaaS where Google CPCs routinely exceed $10-$20. If the anecdotal CPA improvements hold up, the economics could work for high-LTV products. But “anecdotal” is doing a lot of heavy lifting in that sentence, and no one should reallocate meaningful budget based on unverified claims from unnamed beta testers.

Criteo’s 1.5x conversion lift figure is the most interesting data point because it comes from a named partner with a defined sample (500 US retailers). It suggests that users arriving from LLM-generated recommendations may carry higher purchase intent than users from other referral channels. Whether that lift persists as ad volume scales, and as users become habituated to seeing sponsored cards in their ChatGPT conversations, is an open question that no one can answer yet. [1]

OpenAI reportedly hit a $100 million annualized revenue run rate from ads within six weeks of the pilot’s February 9 launch, though actual collected revenue was closer to $8-12 million. [3] That gap between annualized projections and real revenue is worth keeping in mind when evaluating the hype around this platform.

How conversational ad creation performs

OpenAI’s pitch for ChatGPT ads rests on a fundamentally different creative premise than search or social advertising. Because ads appear inline with conversational responses, they inherit a context that traditional display or search ads never have: the user has just asked a specific question and received a specific answer, and the ad appears as a natural extension of that exchange. In theory, this should produce higher relevance and lower friction than a blue link on a SERP or a banner on a content site.

In practice, the creative format is quite constrained. You get a headline, a description, and an optional image. There is no video, no carousel, no interactive element, and no way to customize the ad based on the specific conversation that triggered it. The conversational context is handled entirely by OpenAI’s matching algorithm; the advertiser’s creative is static. This means the “conversational” aspect of these ads is really about placement context, not about the ad unit itself being conversational.

From what I have seen in screenshots shared by beta testers and in reporting from Search Engine Land, the sponsored cards look clean and unobtrusive, which is probably intentional given OpenAI’s stated concern about user experience. [4] Users can dismiss or report ads and disable ad personalization, with data retained for only 30 days. [1] That 30-day retention window is notably shorter than what Google or Meta maintain, which limits OpenAI’s ability to build the kind of longitudinal user profiles that power sophisticated ad targeting on other platforms.

Whether the creative format is “enough” depends entirely on your goals. For brand awareness in a high-attention environment, a simple sponsored card may outperform a display banner that users have trained themselves to ignore. For direct response, the lack of dynamic creative optimization, A/B testing tools, or conversion-optimized formats is a real limitation. You cannot test 15 headline variants against each other the way you would in Google Ads or Meta. You write your copy, upload it, and hope OpenAI’s context matching does the heavy lifting.

Key limitations reported by beta testers

The list of what the ChatGPT Ads Manager cannot do is, at this stage, longer and more instructive than the list of what it can. Beta testers and reporting from multiple outlets have flagged several categories of limitation that range from inconvenient to potentially deal-breaking.

Attribution is the most serious gap. There is no conversion pixel, no postback integration, no connection to GA4 or any third-party measurement platform. Advertisers can see impressions and clicks inside the dashboard, but everything that happens after the click is invisible to the platform. [6] OpenAI has built a separate tool to track whether ChatGPT ads convert, according to Digiday reporting, but it is not yet integrated into the Ads Manager in a way that advertisers can use for real-time optimization. [8] Without attribution, you cannot calculate ROAS, you cannot optimize toward conversions, and you cannot justify spend to a CFO who wants to see pipeline impact.

“If they change the CPM partway through the test, it would mess up how they’re looking at it.”

That quote, from an advertising executive speaking to ppc.land, captures a second frustration: pricing instability. [1] CPMs launched at $60, dropped to $25 in some cases, and CPC bidding appeared for a subset of users at $3-$5, all within a few weeks. For advertisers trying to run controlled tests with consistent variables, a platform that changes its pricing model mid-flight makes rigorous performance evaluation nearly impossible.

Targeting granularity is another weak point. Country-level geo-targeting with no demographic or behavioral segmentation means you cannot isolate high-value audience segments or suppress low-intent traffic. The context hints system is an interesting concept, but it gives advertisers no visibility into how their hints are interpreted, which queries trigger their ads, or why certain impressions were served. There is no search terms report equivalent, no negative keyword functionality, and no way to exclude specific conversation topics beyond OpenAI’s built-in sensitive-topic filters.

Scale is constrained too. With fewer than 20% of eligible users seeing ads on any given day, and with the platform limited to Free/Go tier users in a handful of countries, the available impression volume is a fraction of what Google or Meta can deliver. [3] For brands that need predictable reach at scale, this is not yet a viable primary channel. And the $50,000 minimum spend, while lower than the original $200,000 floor, still prices out most small businesses despite OpenAI’s claim that 80% of pilot advertisers are SMBs. [1]

When to add it to your channel mix

Here is my honest assessment: the ChatGPT Ads Manager is not ready for performance marketers who need measurable, optimizable, attributable results. It is a pre-alpha ad platform wearing a beta label, and the absence of conversion tracking alone should disqualify it from any serious direct-response budget allocation in its current form.

That said, there are specific scenarios where testing makes sense. If you sell a high-LTV product in a category where Google CPCs already exceed $10-$15 (enterprise software, financial services, legal), the $3-$5 CPC range could deliver favorable unit economics even without perfect attribution, provided you have server-side tracking or UTM-based workarounds to measure downstream impact independently. Criteo’s 1.5x conversion lift data, if it holds, suggests that LLM-referred traffic may carry intent signals that justify premium CPCs in the right verticals. [1]

Brand advertisers with awareness objectives and tolerance for CPM-based buying have a clearer use case. A $25-$60 CPM for placement inside a high-attention conversational environment compares favorably to premium programmatic inventory on many publisher sites, and the contextual relevance of appearing alongside a user’s specific question is a genuine differentiator that display networks cannot replicate.

“OpenAI needs to move quickly to establish itself as a legitimate player in the ads business, so launching an ad manager early is an important step.”

Debra Aho Williamson, Sonata Insights [1]

Williamson’s framing is right, but it also reveals the tension at the heart of this launch. OpenAI is shipping an ads platform under IPO pressure, which means the product roadmap is being driven at least partly by investor narrative rather than advertiser readiness. Early advertisers like Target, Adobe, Ford, and Expedia are participating because they can afford to experiment at $50,000 minimums and because being early to a new channel has option value even when the data is thin. [1]

If you are considering a test, the things to watch over the next two quarters are conversion tracking integration (without it, the platform cannot graduate from awareness to performance), search terms transparency (advertisers need to see which conversations trigger their ads), and whether CPC bidding rolls out broadly or stays limited to a handful of testers. Until those three capabilities ship, the ChatGPT Ads Manager is a scouting mission, not a channel strategy. Allocate accordingly: enough budget to learn, not enough to matter if the data never materializes.

Sources

1. OpenAI’s ads manager is live – and the barrier to entry just dropped
2. OpenAI launches ads manager, reduces ChatGPT ad pilot cost to $50,000
3. Quantikal’s Post – LinkedIn
4. Advertisers test ChatGPT Ads Manager – Search Engine Land
5. ChatGPT Ads Now Offer CPC Bidding Between $3 And $5: Report
6. OpenAI adds CPC ads to ChatGPT – Search Engine Land
7. ChatGPT Ads Hit $100M Revenue: Self-Serve April Guide
8. OpenAI builds tool to track whether ChatGPT ads convert – Digiday