OpenAI Ads, Explained: What Brands Need to Know

cresvaOpenAI

Every fifteen or twenty years, the surface where customers see ads moves. Television in the seventies. Search in the early two-thousands. Social in the early twenty-tens. Each time, the brands that recognized the new surface first compounded an advantage that took the next cohort a decade to close. We are at one of those moments again, and the new surface is the conversation itself.

OpenAI is opening ChatGPT as an ad channel. That sentence is short and the implications are not. For roughly eighteen months we have written that the next ad surface would not be a feed, would not be a search results page, would be the answer a model gives when a customer asks for a recommendation. The economics, the user behavior, the platform incentives all pointed at it. The thing that was missing was the platform itself saying yes, we are going to run ads here, and here is how. That is what just happened.

This post is the long version of what brands actually need to know. Not the press release. Not the speculative thread. The version a senior marketer can read once, close the tab, and walk into a Monday meeting with a defensible position. We will cover what OpenAI Ads is, why it matters more than any single channel launch in the last decade, who is actually shopping on ChatGPT, what the integration question looks like inside an existing stack, and what to do in the first ninety days if your brand wants to be on the channel during the window where the inventory is not yet auctioned to death.

What OpenAI Ads actually is

OpenAI Ads is a paid placement system inside ChatGPT. When a user asks ChatGPT a question with commercial intent, the model may surface a sponsored answer or a sponsored recommendation alongside its organic response. The specifics are still being shaped, and OpenAI has been clear that user experience is the constraint, not the variable. That framing matters because it tells you the shape of the inventory before any rate card is published.

Three things follow from that framing. First, ad load will be low by traditional standards. ChatGPT users come for answers, not for ad density, and OpenAI has more to lose from degrading the model than from leaving ad revenue on the table in year one. Second, relevance will be enforced harder than on legacy platforms. The cost of a bad recommendation in a conversation is much higher than the cost of a mistargeted Meta ad, because the bad recommendation breaks the trust that the model is selling against. Third, formats will lean toward the conversational. Banners, interstitials, and full-screen creatives do not fit the surface. Recommendations, answers, and product cards do.

If you are mapping this to existing channels, the closest analogue is not Meta or Google but the early days of Amazon Sponsored Products. There, the ad was structurally indistinguishable from the organic result, the surface was a query, and the brands who learned the auction first compounded position for years. Search ads in 2003 had the same property. The closer the ad gets to looking like the answer, the more the channel rewards brands who deserve the answer.

If you want to see the full product surface inside Cresva today, the OpenAI Ads page covers the workflow. The short version is that ChatGPT becomes a destination next to Meta, Google, and TikTok in the same agent-driven planning loop you already run.

OpenAI rolled out self-serve cost-per-click bidding for ChatGPT ads in April 2026, with click prices typically landing in the three-to-five-dollar range per Digiday's reporting. The Ads Manager beta followed in May, alongside a Conversions API and pixel that AdExchanger covered on launch. Two things follow. First, attribution back to your site is now possible without survey-only methods. Second, the conversion path is explicitly merchant-side: a shift from the September 2025 Instant Checkout pilot that originally let buyers complete transactions inside the conversation. ChatGPT is the discovery surface; your site is where the purchase closes.

Why this is the biggest channel shift since Meta bought Instagram

The Meta acquisition of Instagram in 2012 was a billion-dollar bet that mobile attention would consolidate inside a single graph. The doubters at the time pointed at a thirteen-employee company with no revenue. Inside three years, Instagram ads outperformed Facebook ads on creative attention, audience targeting, and unit economics. The graph that was already there was the asset; the ad layer was layered on top.

ChatGPT is the same kind of asset, with a different shape. The graph is not social, it is conversational. OpenAI disclosed seven hundred million weekly active users when it announced the ad pilot, and roughly fifty million shopping queries a day pass through the conversation. That habit is now nearly two years old in consumer behavior terms, which is approximately the same age the Instagram daily-usage habit had reached when Meta turned on ads at scale. The user base is the asset. The ad layer is what makes it monetizable. The brands who get on early get to learn the channel before the auction matures.

The deeper argument, the one worth holding through the next year of noise, is that this is a shift in kind, not in degree. Meta, Google, and TikTok are all interruption surfaces, the difference between them is the user's mode at the moment of interruption. ChatGPT is not an interruption surface. The user is in intent mode already, and the model is the conduit. That changes the unit economics of the channel because the friction between intent and the merchant site is not three creatives and a landing page, it is one recommendation that earns the click. The buyer leaves ChatGPT and completes the transaction on your site.

We made a related argument in the year brands stop advertising and start answering: the underlying behavior shift is from interrupt to answer. OpenAI Ads is the first scaled monetization layer that sits on top of that shift. Treat it accordingly.

Who ChatGPT shoppers actually are

There is a temptation, with any new channel, to summarize the audience in a single demographic sentence. We will not do that. The honest answer is that ChatGPT shoppers are a meaningfully different population from Meta and Google shoppers, and the population is still resolving in real time. What we can say with confidence is the shape of the behavior, even where the demographic data is thin.

Three behaviors matter more than the audience profile. The first is consolidated research. Buyers who would have opened eight tabs to compare are now asking one question and getting one answer. The substitution rate for browser tabs is high in any product category where the buyer feels uncertain. Skincare, supplements, software, anything where the next ten minutes of research is friction the buyer is happy to outsource.

The second is intent that does not pass through search. A user asking ChatGPT to recommend a moisturizer for sensitive skin is not running that query through Google. The transaction is initiated, lived, and resolved inside the conversation. That intent is real, it is non-trivially sized, and it is invisible to every analytics tool that watches the search bar. The dark-funnel implications are large, and we covered them in detail in the $400B dark funnel post.

The third is replacement. Some categories of repeat purchase, especially low-consideration items, are starting to bypass discovery entirely. The model already knows what the user bought last time. The next recommendation is one sentence, and the buyer lands on the merchant site to repurchase. For brands selling into those categories, the question is not how to win the impression. The question is how to be the brand the model remembers.

None of these three behaviors map cleanly to the personas your team is already targeting on Meta. The user is the same human; the behavior is different. The brands who learn the behavior first will not need to retarget; they will be the default.

What this means for DTC brands

The strategic implications fall into five buckets. We are deliberately not ordering them by impact, because the order depends on your category and your current channel mix.

• The acquisition cost ceiling is going to drop in your category, not rise. New channels behave this way in year one: inventory is undersold, the auction is unsettled, and the brands who learn fast capture asymmetric ROAS until the channel matures. That window typically lasts twelve to eighteen months, sometimes longer.
• Creative discipline will be different. Conversational ad units reward clarity and specificity over emotion. The brands that win will be the ones that can describe what their product does in one sentence that survives being repeated by a model.
• Attribution will get worse before it gets better. Conversion paths that begin inside a ChatGPT conversation will look like direct traffic in your analytics. If you are already on a multi-touch attribution model, you are ahead. If you are on last-click, you will misread the channel and underweight it.
• Brand authority compounds harder. Models are more likely to recommend brands that have signal density, reviews, structured data, expert citations. The work to become recommendable is the same work that pays off across every AI surface, not just ChatGPT.
• Channel concentration risk goes up, then down. Early on, brands that get fluent on OpenAI Ads will have an outsized share of the new channel. Within eighteen months that share will dilute as more brands learn the mechanics. The current window is the asymmetric one.

Two of these five are decisions you can make this quarter. The other three require structural changes to your stack. None of them require you to wait for the channel to mature before acting, which is the point.

The integration question

The temptation with every new channel is to spin up a new tool, a new team, a new dashboard. We have watched DTC brands do this for every channel from TikTok to Pinterest to retail media, and the result is the same every time: the channel becomes an island, the data does not flow, the agents managing one channel cannot see what the agents managing another are doing, and the brand ends up with a portfolio of optimizations that do not compound.

The integration question is not whether to be on OpenAI Ads. It is whether OpenAI Ads will live inside your existing planning loop or become a third or fourth silo that someone manually reconciles on a spreadsheet on Friday. The answer that compounds is the first one. The answer that scales is the first one. The answer most brands will default to is the second one, because adding a channel feels like a software problem and software problems usually get a new tool.

If your Meta, Google, and TikTok activity already lives inside Cresva, ChatGPT becomes a destination your existing agents can plan against. Cresva has early access to OpenAI's Ads Manager beta, so the new channel slots into the same planning loop your team already runs. The same Felix forecasts spend across the new channel, the same Sam runs scenarios that include it, the same Olivia briefs creative that respects the conversational format, the same Dex ships the morning report that includes it next to every other surface. That is the integration argument, and it is not specific to Cresva. The principle is general: one planning loop, one institutional memory, one place where every channel lives. Whatever you build it on, build it once.

The window where OpenAI Ads is undersold and underlearned is the window worth standing inside. Brands that learn the channel during this period will compound advantages that the next cohort will need years to close. Brands that wait for the playbook to mature will be reading the playbook the early movers wrote. If you want to start now, you can try Cresva for free, or read the rest of the cluster to go deeper before you do.