ChatGPT Ads for Agencies: What You Can Run, Measure and Report Today

Two questions keep coming up in client calls: "Should we be running ChatGPT ads?" and "If we do, how will we know it worked?" A year ago the honest answer was that almost nobody could buy them. That has changed fast. OpenAI now sells ads inside ChatGPT through a self-serve Ads Manager, a sales team and a network of partners, and it reports clicks, spend and conversions much like the platforms your clients already know.

The hard part for an agency is not the setup. It is explaining what the numbers mean, how they compare with Google and Meta, and how a paid placement differs from being recommended inside the answer itself. This guide covers what you can run as of early October 2026, what OpenAI actually measures, and how to put the results in a client report without overselling a beta channel.

What are ChatGPT ads, and who can buy them now?

They are sponsored units that appear below a ChatGPT response, chosen by OpenAI's ad system based on the context of the conversation. OpenAI ran a focused pilot with a small group of advertisers from February through April 2026. On May 5, 2026, it announced a beta self-serve Ads Manager, added cost per click bidding next to the original CPM buying, and named agency partners including Dentsu, Omnicom, Publicis and WPP.

Access widened over the summer. On August 31, 2026, OpenAI said self-service was launching across India, Europe, the Middle East and North Africa, and on September 23, 2026 it added Indonesia, Malaysia, the Philippines, Singapore, Thailand, Vietnam and Taiwan. The live country list sits on OpenAI's Ads Manager availability page, and the rule that matters is that the legal entity being billed must be based in a listed country. Some new self-serve accounts can only advertise in their home country at first, until identity verification is approved and they have spent enough there.

Who sees the ads matters as much as who can buy them. OpenAI shows ads only to users on the Free and Go plans. Plus, Pro, Business and Enterprise users do not see them, and neither do accounts identified as belonging to someone under 18. For a B2B client whose buyers mostly use paid workplace plans, that can shrink the reachable audience a lot, so say it before anyone sets a budget.

What can you run: formats, bidding and targeting?

Today you can run one main unit: a text and image ad placed below a ChatGPT answer. According to OpenAI's ad basics guide for advertisers, each ad carries the advertiser name, a favicon, a headline, a description, an image and a landing page. Product feed campaigns are also available for retailers, and OpenAI says more formats are planned.

There are no exact match keywords. At the ad group level you write "context hints" that describe the topics or conversations where the offer is relevant, and OpenAI is clear that hints guide matching but do not guarantee delivery. That is a real change for search specialists used to the keyword and search term control they get in Google Ads campaign reporting.

How does measurement work for ads in ChatGPT?

OpenAI measures delivery inside Ads Manager and conversions through its own Pixel and Conversions API. The standard columns are impressions, clicks, spend, click through rate, average CPC, average CPM and conversions. Retail advertisers who send an order event also get sales and ROAS columns built only from attributed orders.

Conversions are matched to an ad click through a click reference that OpenAI appends to the landing page URL as oppref. The Pixel stores it in a first party cookie, and you can pass it back through the Conversions API from your server. Where a click ID is missing, OpenAI can use hashed contact details (advanced matching) and, where available, modeled conversions, which are included in the reported total.

For agencies, the privacy design shapes what you can promise. OpenAI says advertisers receive aggregated performance insights and never see conversations. You will not get a search term report, a list of the prompts that triggered an ad, or anything resembling the query data in Search Console. OpenAI's own FAQ also says there are no performance benchmarks yet across advertisers or industries, so any "good CTR" figure a client hears is a guess.

How should you report ChatGPT ads results next to Google and Meta?

Report them as a separate paid channel, with the platform's numbers and your analytics numbers shown side by side. Do not blend ChatGPT ads spend into a single "paid search" line, because the auction, matching and attribution are all different, and a blended CPA hides whether the test is working.

1. Tag every landing page URL with UTM parameters. OpenAI confirms static tracking parameters persist on ad clicks. Use one consistent source and medium, then check in GA4 which channel those sessions land in before the first report.
2. Watch for mixed referrals. GA4's AI Assistant channel collects visits whose referrer matches an AI assistant such as ChatGPT. An untagged ad click shares the same domain, so it can blur paid and organic AI traffic.
3. Pull platform metrics from Ads Manager. Export daily values as CSV for impressions, clicks, spend, CPC and conversions.
4. Show both conversion counts with a note. Put Ads Manager conversions next to GA4 key events and state the attribution window used for each.
5. Compare on cost per outcome, not CTR. A click from a decision-stage conversation may be worth more or less than a search click. Cost per lead or return on ad spend settles it.

AgencyReportr does not connect to OpenAI's ad platform, so the spend and impression numbers come from the Ads Manager export. What it can show is the GA4 side: once the ads carry UTM tags, those sessions appear in the campaigns, channels and conversions data of a branded SEO and analytics report, next to the Google and Meta results the client already reviews in Facebook and Instagram ad reports. Keeping the comparison on the same page is what lets a client judge the test fairly.

One more reporting rule: report spend with a lag in mind. OpenAI says spend can be delayed by about 7 to 8 hours while clicks refresh roughly every 15 minutes, so a mid-day screenshot can show clicks with no cost. Pull numbers for complete days only.

ChatGPT ads vs AI visibility: paid placement or organic mention?

They measure two different things, and clients mix them up constantly. An ad is a labeled placement below the answer that you pay for. An organic mention is the answer itself naming or citing the client's brand. OpenAI states that ads are separate from answers and do not influence them, so no budget buys a better recommendation.

That distinction is the most useful thing an agency can explain. A client who shows up in answers for "best accounting software for freelancers" may not need ads on that topic at all, while a client who never appears can use ads to be present while the organic work builds. To know which situation you are in, you need mention data across assistants, which is what tracking brand mentions in AI provides. AgencyReportr reports brand mentions and citations in ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews, alongside an AI assistant traffic section built from GA4.

Put the two side by side in the report: paid results from the ad test, and organic mentions and AI referral visits from the visibility work. Together they answer the client's real question, which is whether people researching their category in ChatGPT end up on their site.

What test plan should you agree with the client?

Agree the rules before launch, because a beta channel with no benchmarks will otherwise be judged on gut feel. A simple plan covers five points:

1. Audience fit: confirm the client's buyers are likely to use free ChatGPT plans in a supported country.
2. Budget and length: set a fixed test budget above the minimum daily budget and a run long enough to collect conversions, given the 24 to 48 hour reporting delay.
3. One success metric: pick cost per lead, cost per sale or ROAS, and the target that counts as a win.
4. Measurement in place: Pixel or Conversions API live, UTM tags on every ad, and attribution windows written down.
5. Reporting cadence: the same section in every report, delivered on the schedule the client already gets through automated branded PDF reports.

Handled this way, ads in ChatGPT become a measured experiment rather than a fashionable line item, and the agency stays the one explaining the numbers instead of defending them.

ChatGPT Ads FAQ

Can any business buy ChatGPT ads now?

Not everywhere. Self-service Ads Manager is open to businesses whose billing legal entity is based in a supported country, which includes the United States, the United Kingdom, India, Canada, Australia and most of Europe. Other advertisers can buy through the OpenAI Ads team or its agency and technology partners. The platform is still labeled beta.

What is the minimum budget for ads in ChatGPT?

OpenAI's help center lists a minimum daily budget by billing currency, for example 25 USD, 15 GBP, 15 EUR or 725 INR. A campaign can use a daily budget or a campaign total budget, and for Clicks campaigns OpenAI suggests starting with a maximum bid of $3 to $5 per click.

Do advertisers see what people typed into ChatGPT?

No. OpenAI says advertisers do not get access to private conversations and receive aggregated performance insights. Ads are matched using the context of the conversation and, where ads personalization is on, select signals from the user's wider ChatGPT use, but the advertiser only sees campaign level reporting.

Why do ChatGPT Ads Manager conversions differ from GA4?

The two tools use different attribution. Ads Manager uses last touch attribution with a 7, 14 or 30 day click window and an optional 1 day view window, and it may include modeled conversions. GA4 uses its own attribution settings and only sees visits it can track. OpenAI states that a difference does not necessarily mean an error.

Does running ads help a brand appear in ChatGPT answers?

No. OpenAI states that ads are clearly labeled, separate from ChatGPT's answers, and do not influence them. Appearing in the answers themselves is earned through content, reputation and citations, and it is measured with AI visibility tracking rather than an ad account.