How to Run ChatGPT Ads: Setup, the $3 Starting Bid, and the Reporting Gap

89 countries, no keywords, and a bid floor near $3

A paid search manager opens ChatGPT Ads Manager for the first time expecting it to look like every other platform they've used, with a keyword planner, an audience builder and a placement tree. What they find is a text box asking them to describe the conversations where the product might be useful, and a warning when their bid goes below $3. The whole setup takes about twenty minutes, and most of that is waiting for a brand review.

That's the honest state of advertising inside ChatGPT in August 2026. It's a real auction with real spend behind it, and it's thinner than the platforms it gets compared to. We walk through the setup end to end, using OpenAI's own advertiser documentation for what the platform does and agency reporting, clearly attributed, for what early campaigns have actually paid. If you want the wider picture of what advertising inside ChatGPT is, start there. This piece assumes you're past that and about to launch something.

What you are actually buying when you buy a ChatGPT ad

Before you touch a setting, it helps to be precise about what the unit is. It isn't a search result or a feed placement, and the differences change what your creative has to do.

The unit is six parts, and four of them are yours

According to OpenAI's advertiser documentation, "Ads appear below ChatGPT responses." Each ad is assembled from six declared parts:

• Advertiser name, from your account
• Favicon, your logo, uploaded once
• Title, the headline, yours to write, though OpenAI may create new or updated versions unless creative enhancements are off
• Copy, the description, also yours to write, though OpenAI may update or translate it unless that setting is off
• Landing page, the destination, which OpenAI may change to another page on your site unless that setting is off
• Image asset, the creative, which OpenAI may resize to fit the ad format unless that setting is off. Since October 26, 2026, enabled accounts can also create ads with four to six square images

The first two come from account setup and never change. Everything a reader actually looks at is in the last four. You supply all four, and OpenAI can adapt them unless creative enhancements are off.

Those fields are tight. In the Ads API, a creative is either a chat_card or a product_ad_template, the title accepts 3 to 50 characters, and the body accepts up to 100. There's no room to explain, so the image carries unusual weight.

Who is on the other side of it

The audience is limited by plan and by age. OpenAI is explicit about who's left out: "We do not show ads to users on Plus, Pro, or any Business plan, or to any users who tell us or we predict are under the age of 18." So the people you can reach are adults on the lower-tier consumer plans, and when they see your ad, they're actively asking questions rather than scrolling.

That matters more for choosing an angle than for targeting. Someone who's just asked a comparison question is further down the funnel than someone who stopped scrolling on a video, and that's a different creative job from the one ecommerce video ads usually do.

The auction is relevance-weighted and second-price

OpenAI spells out the mechanism: "We use a relevance-weighted, second-price auction to select among eligible ads." Relevance works as a multiplier on your bid here, so it does more than break ties. OpenAI also lists the relevance inputs, saying the system "considers multiple signals, including the context and intent of the current conversation, the ad's landing page, title, copy, advertiser-provided context hints."

Four of the five signals are things you control, and the fifth is the conversation itself. That control now comes with a condition, because OpenAI can adapt the title, the copy and the landing page unless creative enhancements are off. On a platform with no keyword bidding, your copy carries the message and is also part of how you qualify for the auction at all.

Get the account cleared before you plan the campaign

Two setup steps gate everything else, and both are easy to miss until late.

Brand review is a gate, not a formality

OpenAI's partner setup documentation puts it bluntly: "An account cannot serve ads until its brand review is approved." OpenAI describes reviews as typically taking only a few minutes, and you can check the account's review_status field to see where yours stands. One common rejection reason is a missing favicon, so fix that before you build anything.

Plan for review to come first instead of running it in parallel with your launch. A campaign that's built, funded and scheduled on an unapproved account simply doesn't deliver. The pre-flight list is short:

• Brand review approved, confirmed on the account's review_status
• A favicon uploaded and assigned, since a missing one is a named rejection reason
• Conversion tracking live if you plan to use the conversions goal
• Tracking parameters appended to every landing page URL
• Creative enhancements decided at the account level and per campaign, since they're on by default for self-serve campaigns

ChatGPT Ads is available in 89 countries

OpenAI said on October 21, 2026, that it was expanding ChatGPT Ads to 26 more countries, "extending availability to 89 countries worldwide." The additions include 14 more Latin American countries, among them Argentina, Chile, Colombia and Peru. OpenAI says advertisers in the new countries can go through technology partners or use self-service access in Ads Manager.

Whether you can buy self-serve depends on where the business that's billed is based. We can't open OpenAI's availability page, but its English copy archived on October 12 says "the legal entity that will advertise and be billed must be based in a country listed as available," and that you can still sign up for an account where self-serve hasn't opened and hear from OpenAI when it does. That copy lists 63 countries. The newest archived copy, the page's Spanish-language version from October 23, is an automatic translation that says the same two things and lists 89 countries, each marked available for self-service.

That list has changed quickly. When we first published this guide, the page's August 3 copy listed seven countries and said availability "may continue to evolve as testing expands." So don't bake any country list into a plan, ours included, and check the page for your billing country on the day you build the media plan.

Pick the goal, and know what it bills

A campaign takes one of three bidding types. The API reference lists impressions, clicks or conversions, with the note "Defaults to impressions", which is worth knowing because most advertisers don't want the default.

Goal	You optimize for	You pay per
Impressions (CPM)	Reach and awareness	1,000 impressions
Clicks (CPC)	Traffic	Valid click
Conversions (oCPC)	A tracked action	Valid click

The third row is the one that trips people up. OpenAI's documentation puts it plainly: "Billing does not change to pay per conversion. OpenAI charges you only when a valid click occurs, and the auction determines the actual CPC." The conversions goal changes which clicks the auction goes after, but you're still billed per valid click. We covered that in detail in how ChatGPT Ads conversion bidding works, including the prerequisites that have to be live first.

Creative enhancements are on by default, so decide before you launch

Since Friday, October 30, 2026, an ad in a self-serve campaign can run in versions OpenAI generated, unless you've opted out. OpenAI's October 20 email to advertisers, which Search Engine Roundtable published, says "we'll enable creative enhancements for your new and existing self-serve campaigns." OpenAI's help page, as the Bangkok agency Relevant Audience quoted it on October 31, leaves out the word self-serve and says creative enhancements "will be enabled automatically for new and existing campaigns." We couldn't open that page ourselves, so its wording in this section comes from the reports that quote it.

PPC Land noted on October 21 that the help page doesn't say whether campaigns bought through OpenAI's sales team or technology partners are included, and we haven't found an answer since. It's a default, and it doesn't reach every campaign, because OpenAI's email says existing account and campaign opt-outs are preserved. The same email says a campaign's own setting takes priority over the account default.

One setting covers four parts of your ad

The help text that Search Engine Roundtable reproduced lists four kinds of change, and Search Engine Journal reports that OpenAI's page describes one setting for all of them.

• Headlines and descriptions. "OpenAI may create new or updated text to connect your offering to the conversation or user interests."
• Language. "OpenAI may translate your ad text into another language."
• Images. "Images may be resized to fit the ad format."
• Destination URL. "The destination URL may be optimized to direct people to a relevant page on your website."

So your ad's text, language, image size and landing page can each run in a version OpenAI generated. OpenAI's email says that when the setting is enabled, these ads "can run automatically without requiring you to review or approve each version," and Relevant Audience quotes the page's FAQ as saying the experience "does not provide a preview of every variation."

That's the default as OpenAI describes it. The only before and after we've seen is an illustration from its help page, as Relevant Audience reported it on October 23. In a conversation about planning quick weeknight dinners, "Fresh groceries delivered" could become "Groceries for quick weeknight dinners." OpenAI notes that the example assumes the original ad and landing page support the details shown, and that "actual wording may vary." We haven't found a report from anyone who saw a changed ad after October 30, so we can't tell you how often it happens or how far a version moves from what you wrote.

Where to switch it off

The account setting is in Ads Manager (Beta), under Settings, on a card labeled Creative enhancement (singular on the card, plural in OpenAI's email). Search Engine Roundtable's October 20 screenshot shows it there, switched on. OpenAI's email says you can change your settings "at any time in Ads Manager," and the link in it reads Manage settings.

A campaign's own setting overrides that card, so you can leave the account on and switch one campaign off, or the other way around. No page we could read shows where the campaign setting is. In September, the ad software company Sprites reported a Text customization switch in the Advanced section of campaign creation. We don't know if it's the same setting, but we'd look there first.

If you build through OpenAI's Ads API, we can't tell you how to control this. None of the four reports we read (Search Engine Journal, Search Engine Roundtable, PPC Land and Relevant Audience) says whether you can switch it off through the API. We read OpenAI's public Ads API spec and its Ads docs on October 31, 2026, and found no field for it and no mention of it. That doesn't prove there's no way to do it. We'd ask OpenAI whether the account setting applies to campaigns you build through the API.

OpenAI's terms describe how it works. In a copy of its Ad Tools Terms archived on October 22, section 3.1 says Creative Tools may "make Generated Creatives available for your review or automatically apply Generated Creatives to your Campaign in accordance with your Campaign settings," and that you "can manage or disable Creative Tools through applicable Ads Manager controls." Neither sentence says what the default is or gives a date.

What we'd do before the first campaign

We'd decide this per account before the first campaign, then confirm it on each campaign before launch. In a test, we'd keep it the same in every cell, because you can't compare your creative when one side may be running OpenAI's version. If the test is about your own creative, we'd switch it off in every cell, the same call we make for Meta's enhancements in a creative test.

OpenAI's page, as Search Engine Journal and Relevant Audience quote it, says "Changes over time do not, on their own, establish the effect of creative enhancements." We read that as a warning against judging the setting on a before and after.

Your UTMs need a look too. Search Engine Journal and Relevant Audience both say OpenAI's pages don't explain how tracking parameters carry over if the destination changes. We'd check your own analytics to make sure ChatGPT sessions still arrive tagged.

If you leave the setting on, we'd keep your landing pages current. OpenAI's email says it uses your "existing ad and landing-page content," and section 3.2 of those archived terms makes you responsible for generated versions applied to your campaign, including the accuracy of what they say about pricing and availability.

What OpenAI hasn't said

OpenAI hasn't said whether creative enhancements apply to multi-image ads. Its developer changelog says that since October 26, 2026, enabled accounts can create ads with four to six square images. Search Engine Roundtable reported on October 30 that the format is in open beta for all advertisers, and Search Engine Journal found no indication either way on enhancements when it checked that day.

Nor has OpenAI said how the setting works in the EEA. Its docs say personalized ads are "not yet available" there, and we found nothing on what that means for creative enhancements.

Text changes may have been on by default before now, and the record conflicts. OpenAI's September 16 post described text customization as something you opt into, while Sprites reported that the switch was on by default in its own account on September 13. Search Engine Journal said on October 30 that the help page doesn't explain how creative enhancements relate to that September option. What looks new to us is the reach into existing campaigns, plus one setting that also covers image resizing and destination URLs. We'd check the setting even if you turned text customization off in September.

Targeting is three dials, and one of them is a paragraph

This is where the platform differs most from Google Ads, and it's where operators get confused most often.

Geography: country, region, market, looked up live

OpenAI's location targeting docs, as we read them on October 31, 2026, let you "choose the countries, regions, or markets where your ads can deliver." If you don't set any location targeting, "the campaign can target all available locations," and that's a default worth changing deliberately.

There are three levels, and they nest the way a media planner would expect:

• Country, the default unit, and the only one that matches the reporting breakdown
• Region, for state- or province-level control
• Market, for metro-level control, such as New York in OpenAI's own example

OpenAI calls these markets, and its docs don't mention DMAs, so look up the metro you want instead of assuming it matches a DMA you buy on other platforms. You can also exclude locations. OpenAI's own example targets the US and leaves out California, and its docs allow up to 2,500 IDs in those include and exclude lists.

The set of targetable locations isn't a published list you can memorize. The docs point you to a lookup that finds "locations currently available for targeting," and they link to a CSV download of the current location catalog. OpenAI adds that a valid country code doesn't mean the country is "available for advertising in your account," so look locations up each time you build a campaign instead of reusing an old list.

Context hints: a description, not a keyword list

Context hints replace keyword targeting, and they really are different. OpenAI describes them this way:

"At the ad group level, advertisers can provide context hints that describe the conversations, topics, or keywords where their products or services may be relevant. These hints help guide ad matching, but they are not exact-match keywords and do not guarantee delivery in specific conversations."

That has three practical consequences:

• An ad group now stands for a theme instead of a match type. Hints are set at the ad group level, so your account structure should split by topic rather than by keyword variant.
• Breadth is cheap. OpenAI documents the field as "Free-form audience or placement hints," and an ad group accepts up to 2,000 of them.
• You can't prune your way to precision. Because nothing is exact-match, there's no search query report to mine and no negative keyword to add when delivery drifts.

Custom audiences and the 25,000-user floor

You can use first-party lists, but they come with a hard size gate. In OpenAI's words, "Note that each custom audience must include at least 25,000 matched users before it can be used. We recommend audiences of at least 100,000 users."

Two more constraints shape how you use them. "Custom audiences cannot be edited after creation," so refreshing a list means creating a new audience and archiving the old one. And "Ads Manager does not show individual matched users or let you select specific people from an audience," so you can switch an audience on or off, but you can't look inside it. For most advertisers below enterprise scale, this dial simply isn't available.

What to bid, and what the first campaigns actually paid

Two numbers get passed around here, and people confuse them all the time, even though they're different kinds of fact. One is a recommendation from OpenAI. The other is an observation from an advertiser.

OpenAI's own starting number

OpenAI publishes its recommendation: "For CPC campaigns, we recommend starting with a maximum bid of $3-$5 USD per click." And it notes that "Ads Manager may also provide bid-strength guidance to indicate whether a bid is likely to be competitive or may limit delivery."

That guidance is the warning our paid search manager ran into at the top of this piece. It tells you a bid is unlikely to clear a relevance-weighted auction, but it isn't a hard floor.

What one agency reports paying

For realized cost, the clearest dated figure comes from practitioner reporting. Writing in Search Engine Land on August 10, 2026, StubGroup CEO John Horn says of his agency's accounts, "We're seeing average CPCs in the $2-$5 range across industries."

If you repeat that figure, attribute it correctly. It's one agency's account data across its own client mix, and it isn't an OpenAI rate card or an audited benchmark. It's a useful sanity range for a first budget, but it's no use as a target.

Why the bid is doing more work than it looks

Because the auction is relevance-weighted, your bid and your creative aren't independent levers. A stronger ad wins impressions that a weaker ad would need a higher bid to buy. So before you raise a bid to fix delivery, check whether the ad is what's losing. Raising a bid on a weak creative buys volume at the worst possible price, the same trap that chasing click-through rate as a success metric sets on every other channel.

Reporting names the winning ad, never the winning conversation

Measurement is the part of this platform most likely to surprise someone coming from Google Ads, so it's worth being exact about it without getting dramatic.

The metrics you get

OpenAI lists the set, and you should read it as complete rather than as a sample:

• Impressions
• Clicks
• Spend
• Click-through rate
• Average CPC
• Average CPM
• Conversions

Conversions need tracking to be live, through either the pixel or the server-side route. That pixel's default also changed in August, when OpenAI set automatic advanced matching to turn on for existing web pixels, which we covered in the automatic advanced matching switch.

You can also layer on your own tracking. OpenAI says static tracking parameters such as UTMs "persist on ad clicks and can be used in existing analytics tools to understand traffic from ChatGPT Ads." For a first campaign, that's the most valuable ten minutes of setup you'll spend, because it moves the analysis into a tool you already trust. Search Engine Journal and Relevant Audience report that OpenAI's pages don't explain how those parameters carry over if creative enhancements change the destination. We'd check that in your own analytics.

One breakdown dimension, and rows down to the ad

The API's changelog records what segmentation exists, in an entry that reads "Added segmented insights for product, country, and device breakdowns." You get one optional dimension per request, by product, country or device. Rows can be aggregated down to the individual ad, so per-creative performance really is available.

That last point matters, and it often gets overstated in the other direction. You can see which ad won, but not why.

The gap that matters

Conversation context and intent are named signals in ad selection, but they appear nowhere in the reporting. Nothing in Ads Manager or the insights segments describes the chat that produced an impression, so the single richest input into the auction is the one you never get back.

That's the structural fact to design around. On Google Ads, a search terms report tells you which query bought the click, and you respond by editing the query set. Here, there's no query set to edit and no report to edit it from, so you close the feedback loop on the ad instead of on the audience.

One advertiser's numbers, and how far they stretch

A performance figure is making the rounds right now, and any honest guide has to deal with it, including the way it's being over-read.

What the comparison showed

In the same August 10 guide, Horn published GA4 data covering roughly 1,500 users who reached one website through Google Ads and through ChatGPT Ads. In that sample, Google Ads traffic converted at 3.71% and ChatGPT Ads traffic at 0.21%. Average engagement time was 41 seconds versus 17 seconds.

Why 1,500 users is not a verdict

That's one advertiser, one site and one period, and the sample is small enough that a handful of orders moves the rate. It tells you about a single account rather than measuring the platform, and the sample size belongs in every sentence that repeats it.

The directional read that survives is softer and more useful: traffic arriving mid-conversation lands colder than traffic arriving from a typed commercial query, so landing pages tuned for high-intent search may under-serve it. That is a hypothesis to test against your own numbers It doesn't justify skipping the channel, and you shouldn't plan against it as a benchmark.

So run your own comparison, give it enough volume to mean something, and set up the first two weeks so the result holds up:

• Isolate the traffic. Tag every landing page URL so you can separate ChatGPT Ads sessions in your own analytics, and not just in Ads Manager.
• Hold the landing page constant. Use one destination across all ads, so the creative is the variable you're testing and the page stays fixed.
• Hold creative enhancements constant. Set them the same way in every test cell. With them on, OpenAI may change the destination you just locked in, and you can't compare your creative when one side may be running OpenAI's version.
• Start at the recommended bid. Begin in the $3 to $5 band, and change it only after the creative has had a fair read.
• Run several distinct images instead of several headlines. The image is the biggest surface in the unit and the cheapest thing to vary.
• Judge on the ad-level rows. That's the one clean signal the platform gives you, so let it decide what survives. With creative enhancements on, a row can include versions you never saw, and Relevant Audience reports that OpenAI's pages describe no report listing each enhanced version. We'd read that row as your ad plus whatever changes OpenAI made.

Our guidance on how many ad creatives a test actually needs applies here more than usual, because the platform gives you fewer other things to vary, and structured creative testing is what keeps a thin-data channel honest.

How Novoads solves the creative-supply problem ChatGPT Ads creates

Add up the constraints and you get a platform where:

• Targeting is geography plus a description you can't verify
• Bidding is one number, multiplied by a relevance score you never see
• Reporting shows you the winning ad without explaining why it won

Of everything you can change, the creative is the dial with the most range, and the image is the biggest piece of it, because a 50-character title and a 100-character body leave the picture to carry the argument. It's the same conclusion creative analytics keeps reaching on mature channels, and here you reach it early because the alternatives were never built.

That makes creative supply the bottleneck. Testing a real spread of images means producing a real spread of images, and that's exactly the step where a small team runs out of budget. Novoads takes an uploaded product photo and generates ad images through GPT Image 2 at a fixed 1 credit per image, so you can put eight distinct treatments of one product into an ad group without booking a shoot. The same product image can also become the UGC video ads you run on channels where video is the unit, so one asset library feeds several placements.

Novoads starts at $15/month (Starter, 200 credits per month). Plus is $49/month (1,000 credits). For more volume, Ultra starts at $129/month (3,000 credits). You can cancel anytime.

Creative is the only dial with real range

Every young ad platform starts out thin, and the thin parts show you where the leverage is. With ChatGPT Ads, OpenAI has taken away the keyword, hidden the conversation and capped the copy at a sentence. The picture and the promise stay entirely in your hands, but only while creative enhancements are off. On a channel that won't tell you which conversation bought the click, the fastest way to learn is to put more distinct creative into the auction and read the scoreboard it does give you. That makes production throughput the real limit, so it's worth knowing the number. The median time for a finished video clip, across 1,045 measured renders, was between two and seven minutes, depending on the model. The mirror image is playing out on Google's side of the market, where the new Ask Advisor agents across Ads and Analytics make the measurement half of the job cheaper without touching the creative half at all. Advertisers who treat creative production as the main job, instead of waiting for targeting controls to arrive, will have a year of learning before everyone else starts.