ChatGPT Ads Manager, the 2026 advertiser guide to OpenAI's new platform

On May 5, 2026, OpenAI flipped the ChatGPT Ads Manager from a curated pilot with a $50,000 spend minimum to a self-serve product any U.S. business can sign up for in about five minutes. That's the headline. The interesting story is everything underneath it, because what's actually live at ads.openai.com right now is a brand-new ad platform with one ad format, country-level geo targeting, no audience segmentation, and a CPC bid range that lands somewhere between LinkedIn and a high-intent Google search keyword.

We've spent the last few days inside the new interface, against the backdrop of $550M plus in ad spend we manage across Google, Meta, programmatic, and LinkedIn. This post is the working advertiser's guide to the ChatGPT Ads Manager. What it is, what it costs, what it can and can't do, how it compares to the platforms you already run, and what an agency or a small business should actually do about it in the next 90 days.

If you came here looking for a five-tool stack of breathless takes about how this changes everything, you're in the wrong place. This is the practical version. The platform is real, the placement is genuinely new, and the limitations are also real. We'll cover all of it.

What is the ChatGPT Ads Manager?

The ChatGPT Ads Manager is OpenAI's first self-serve advertising platform. It lives at ads.openai.com and runs ads inside the ChatGPT product, which OpenAI says now reaches around 800 million weekly users. The interface uses the same three-tier hierarchy a paid media buyer already knows from Google Ads. Campaigns at the top, ad groups underneath, and individual ads at the bottom. Anyone who's ever shipped a Google Ads search campaign can navigate it on the first try.

The platform supports two campaign objectives at launch, Reach (CPM-based) and Clicks (CPC-based). Reporting includes impressions, clicks, spend, CTR, average CPC, average CPM, and conversions. Conversions are tracked through OpenAI's own pixel plus a server-side Conversions API, mirroring the structure Meta and TikTok use. Reporting is aggregated, which means you can see how a campaign performed but not which conversation a specific click came from.

What makes the ChatGPT advertising platform genuinely different from Google or Meta is the placement itself. These aren't ads against keywords, audiences, or feed scrolls. They're ads served inside a live AI conversation where someone is comparing options, learning about a category, or making a decision. OpenAI has been transparent about that, framing the value prop as advertising in "active and decision-oriented" conversations rather than passive scroll surfaces. Whether that lives up to the promise is going to take a couple of quarters of real data to know.

How ChatGPT ads actually work

The mechanic that's most foreign to anyone coming from Google Ads is targeting. There are no keywords. There are no audiences. There's no remarketing pixel pool. Instead, advertisers write what OpenAI calls "context hints," which are plain-language descriptions of the conversations where the ad should appear. Something like, "Users comparing accounting software for service businesses with under 50 employees." The model reads the in-progress chat, evaluates it against the context hint, the headline, the description, and the landing page, and decides whether the ad is a relevant fit.

Underneath that sits a relevance-weighted, second-price auction. Every eligible conversation triggers an auction across the advertisers whose context hints the model thinks match. Each ad gets an effective bid that combines the max CPC or CPM with a relevance score. The highest-effective-bid ad wins and pays one cent more than the second-highest bid, the same general structure Google uses for keyword auctions. The output is that bidding above market doesn't guarantee a placement if your relevance is weak, and bidding below market can still win if your relevance is unusually strong.

The other thing worth understanding is that targeting is currently country-level only. There's no state, no DMA, no city, and no radius. According to coverage from Digiday's reporting on the launch, OpenAI's Ads and Monetization Lead Asad Awan acknowledged the company is working on more granular geo, third-party measurement, and CPA bidding, but didn't commit to a timeline. For local businesses, that's the most consequential limitation in the platform today.

The onboarding process, step by step

The signup flow takes about five minutes if your business documents are already in front of you. Verification then takes a few business days, and OpenAI has warned that timelines may stretch as demand spikes around the public launch.

1. Account creation. Visit ads.openai.com and sign up with a business email. The platform isn't tied to a personal ChatGPT account.
2. Business information. Enter your legal business name, EIN, and physical business address. This is a hard requirement, not optional.
3. Category selection. Pick a primary advertising category from the eligible list (household goods, local services, travel and entertainment, digital products, or education at launch).
4. Payment method. Add a credit card or ACH. There's no spend minimum but you must have a valid payment method before campaigns can serve.
5. Pixel and Conversions API setup. Install the OpenAI ads pixel on your site, or wire up the server-side Conversions API. We strongly recommend doing both, the same way we run Meta and TikTok.
6. Verification. OpenAI reviews the application. Approval typically arrives in two to four business days, though some advertisers in the first week of public launch reported longer waits.
7. First campaign build. Choose Reach or Clicks. Set a budget. Write your context hints, headline, description, and landing page. Set your max bid. Launch.

For agencies, the same flow applies, but you'll want a single agency-managed parent account that can be tied to multiple advertiser entities as the platform expands manager-account functionality.

Ad formats and targeting at launch

The ad format is deliberately small. One favicon, one headline, and a short description. There's no rich media, no video, no carousel, no product feed, no shopping unit. Visually, it sits inside or alongside a ChatGPT response and looks closer to a sponsored search result than to a Meta or TikTok feed ad. The advertiser's URL drives the click.

Targeting is the area where the platform is furthest behind every other major channel. Here's what's available, and what isn't, at launch.

• Country-level geo targeting (U.S. only, no state, DMA, city, or radius)
• Contextual targeting through plain-language context hints
• Category-level audience signal through OpenAI's eligibility taxonomy
• Frequency capping at the campaign level (basic)
• Day-of-week and time-of-day pacing through campaign budgets

• No demographic targeting (age, gender, household income)
• No interest or affinity audiences
• No first-party customer match or list-based audiences
• No retargeting or lookalike audiences
• No state, DMA, city, or radius geo targeting
• No keyword targeting in the Google sense

If you've spent any time inside a programmatic DSP or running Meta's audience tools, the absence of audience targeting is the part that takes the most adjustment. Relevance is doing all the work that audience signals do on Meta. That's a feature, not a bug, but it requires writing context hints with the same care a search advertiser writes ad copy.

Pricing, bidding models, and what to expect to pay

There's no spend minimum on the self-serve platform. Reach and Clicks are the two campaign objectives. The bidding inside each is straightforward.

For practical budget sizing, here's how we're thinking about it. According to Search Engine Journal's reporting on screenshots from inside the platform, $3 to $5 CPC bids land roughly between non-brand Google search and LinkedIn for cost per click. With a 1% to 3% CTR (which is what early advertisers are seeing in early case studies), a $1,000 test budget gets you somewhere between 10,000 and 33,000 impressions and 100 to 500 clicks. That's enough to learn the auction, not enough to draw a strong conclusion about ROAS or CPA.

Here's how we'd size a test by advertiser type.

• Small business test: $500 to $2,000 over two to three weeks. Two campaigns. One ad group each. Two ad variations each.
• Mid-market test: $5,000 to $15,000 over four weeks. Three to five campaigns mapped to your top product categories. Three to five context hint variations per ad group.
• Enterprise pilot: $25,000 plus per month, ideally run alongside a managed engagement through one of OpenAI's holding-company partners (Dentsu, Omnicom, Publicis, WPP) for early-access features.

ChatGPT Ads vs Google Ads vs Meta Ads

The cleanest way to think about ChatGPT Ads in 2026 is that it's a new placement, not a new channel architecture. It doesn't replace search intent or feed-based social. It sits next to them and reaches people in a moment those platforms can't.

Where ChatGPT Ads has a real edge is the moment of placement. People who are deep in a chat about which CRM to buy, what laptop fits their budget, or how to plan a trip to Lisbon are sitting in a higher-intent decision context than most feed surfaces can deliver. Where it's behind is on every measurement and targeting axis paid media buyers have built workflows around for the last decade. The smart way to enter is to assume the placement is real and the operational stack around it isn't yet.

Advantages and limitations during the public rollout

The early-access window has its own dynamics. Some are real opportunities. Others are reasons to wait.

What works in the advertiser's favor right now

• No spend minimum after the $50,000 pilot phase ended, which means a $500 test is fine.
• CPMs falling from $60 to roughly $25 over the first ten weeks suggests inventory is expanding faster than demand, which is good for early entrants.
• The ad format is so simple that creative production cost is near zero. A favicon, a headline, a description, and a landing page is the entire creative ask.
• Competition inside the auction is still light in most categories, especially outside ecommerce.
• The platform's auction rewards relevance heavily, which gives small advertisers with sharp context hints a real shot against bigger brands with bigger budgets.

What still works against the advertiser

• Country-level geo only kills the platform for local service businesses outside niche categories.
• No audience segmentation means you can't apply the customer-match or lookalike playbooks that drive most modern Meta and Google performance work.
• Aggregated reporting limits how much you can learn from a single campaign about which kinds of conversations actually convert.
• No CPA bidding yet means you're optimizing for clicks while measuring against conversions, which is harder.
• Inventory is still uneven. Pilot advertisers reported struggling to fully exhaust budgets due to fill rate.
• No third-party measurement means you're trusting OpenAI's own attribution math, which the broader industry hasn't independently validated yet.

Early performance data and advertiser sentiment

Hard performance numbers are still thin. The pilot ran for several months under NDA with a small advertiser cohort, and most public benchmarks are based on screenshots and reporter outreach rather than published case studies. What we can say from public coverage and our own conversations with peer agencies is roughly this.

CTRs in early campaigns are running 1% to 3% on average, with stronger performance reported by advertisers whose context hints map closely to high-intent decision conversations. CPMs that started at the $60 default have settled in the $20 to $35 range as inventory has expanded. CPCs are landing close to the $3 to $5 recommended max, which puts ChatGPT Ads in the same general band as B2B LinkedIn or non-brand Google search.

Agency sentiment is cautiously positive. Debra Aho Williamson of Sonata Insights, quoted in Adweek's launch coverage, framed the platform as "demonstrating that it understands the basic building blocks that are necessary for advertisers to feel comfortable testing on ChatGPT." That maps to what we're hearing internally. The infrastructure is real. The reach is real. The optimization stack still needs the next 12 months of work to be competitive with the platforms advertisers run today.

What agencies and small businesses should know before adopting

Our actual recommendation, after a week inside the new interface and a year of watching the pilot from the sidelines, is segmented by who's asking.

For agencies

Stand up a manager account this week. Get the pixel and Conversions API documented for client onboarding. Build a context-hint writing playbook the same way you'd build a Google Ads keyword research playbook, because that's the closest analog. Don't sell ChatGPT Ads as a primary channel to clients yet. Sell it as a test layer running alongside an existing program, with an explicit budget cap, an explicit measurement window, and an explicit go/no-go criterion at the end of the test. The risk to the agency relationship if you over-promise on a platform this immature is much higher than the upside of being first.

For small businesses

Claim the account. Run a small test if your category is eligible and your geography isn't local-only. Treat anything you spend in the next 90 days as an information cost, not a performance cost. The thing you're buying with that money isn't conversions. It's an early read on whether ChatGPT Ads will become a meaningful channel for your business as the platform matures. If you're a local plumber or a city-radius restaurant, skip it for now and revisit when state or DMA targeting ships.

For ecommerce and direct-to-consumer brands

This is the strongest fit at launch. Decision-oriented conversations about products, comparisons, gift ideas, and category research map directly to the placement. We'd run a test budget of $5,000 to $15,000 across three to five context hints, measure against a 30-day window, and benchmark against a matched-spend Meta or Google campaign. Comparable performance plus incremental reach is the win condition.

For B2B and SaaS

The auction model and intent context favor B2B more than the format size suggests. Buyers researching software inside ChatGPT are doing real evaluation work. The ad format is small enough that strong category-defining headlines and disciplined landing pages do most of the heavy lifting. We'd test against the same audiences a LinkedIn or non-brand Google campaign serves and compare CPL head to head.

The bottom line