AEDAnswer Engine Daily
← All stories

Explained: How are ChatGPT ads performing? OpenAI’s case studies show strong returns, but it has not published a click-through rate

Sandeep Amar · November 17, 2026 · 10 min read
Explained: How are ChatGPT ads performing? OpenAI’s case studies show strong returns, but it has not published a click-through rate
AEO · Answer Engine Daily

ChatGPT ads reached $1 billion in annualized revenue run rate “in less than 200 days after launch,” OpenAI said on August 31, 2026, and “tens of thousands of advertisers” now use the platform. On the same day, self-serve buying opened in India, Europe, the Middle East and North Africa. When checked on November 16, OpenAI’s Ads Manager availability page listed 89 countries and territories for self-serve access, and the ads homepage offered new Indian advertisers ₹5,000 in credit for spending ₹5,000, until December 31.

However, OpenAI has not published a platform-wide click-through rate (CTR), cost per click (CPC), cost per thousand impressions (CPM) or conversion rate. What it has published are seven customer stories, each chosen by OpenAI, each using a different metric, time window and attribution method. The only absolute CTR figures available come from companies outside OpenAI, which posted their own results on X.

This explainer puts those numbers side by side, says what each one measures, and what none of them can tell you.

What are ChatGPT ads?

Sponsored placements shown to logged-in adults on ChatGPT’s Free and Go plans. Plus, Pro, Business, Enterprise and Education plans are ad-free, according to OpenAI’s testing post. The first ads appeared “at the bottom of answers in ChatGPT when there’s a relevant sponsored product or service based on your current conversation,” OpenAI said in January.

Targeting is not keyword-based. OpenAI says it matches ads “with the topic of your conversation, your past chats, and past interactions with ads.” Advertisers describe the conversations they want to appear in using what OpenAI calls context hints. Ads are not shown to users OpenAI predicts are under 18, or “near sensitive or regulated topics like health, mental health or politics,” the testing post says.

OpenAI’s core promise to users is that “Ads do not influence the answers ChatGPT gives you.” Advertisers “only receive aggregate information about how their ads perform such as number of views or clicks.”

How has the platform grown?

Quickly, by OpenAI’s own account:

  • Pricing: The pilot sold ads only on a CPM basis. CPC bidding arrived on May 5, and conversion optimisation followed. By August 31, “CPC and outcome-optimized bidding now account for the majority of campaigns,” OpenAI said.
  • Buying: Ads were first sold directly to “a small group of advertisers,” then through agency partners Dentsu, Omnicom, Publicis and WPP, and technology partners including Adobe, Criteo and StackAdapt. A self-serve Ads Manager opened in beta on May 5.
  • Advertisers: “Tens of thousands of marketers” by August 18. Small and medium businesses “now represent a material share of the business,” OpenAI said on August 31, without giving a figure.
  • Markets: Over 40 countries on August 31 and “more than 60” by September 23.
  • Revenue: $1 billion annualized run rate. OpenAI did not say which date it counted its “less than 200 days” from. February 9, when the US test began, was 203 days before the August 31 post, by our calculation.

What results has OpenAI published?

Seven case studies on its customer stories page, plus partner findings in an October 5 measurement post. The headline numbers:

  • Loop Earplugs (consumer goods): 5.1x return on ad spend (ROAS) and a 3.7% post-click conversion rate across five product-feed campaigns, and 1.8x ROAS on a non-feed EU campaign. Window: August 18 to September 8, 2026.
  • Newegg (electronics retail): 3x ROAS across campaigns over 28 days, and 7x during a two-week sale campaign. No dates given.
  • Talkpal (language learning app): About 26% higher ROAS and about 51% lower cost per trial activation than “an established native advertising channel.” Based on Talkpal’s “internal estimates.”
  • Angi (home services): ROAS up more than 38% from Q2 to Q3, and cost per lead down 34% since it launched conversion-optimised campaigns. No absolute ROAS or cost per lead given.
  • Jotform (online forms): More than 10% of clicks became signups. Cost per signup about 30% lower than paid search and 60% lower than paid social.
  • LegalNature (legal documents): Nearly 40% of visitors started a document and about 8% completed one. Window: July 1 to August 23, 2026.
  • Stream (developer APIs): Roughly 2x ROAS on one campaign, and more than 70% of clicks became website sessions, against “below 10%” for unnamed channels it was tested against. Window: June 5 to August 5.
  • WeightWatchers: Attributed cost per acquisition 15.3% below its blended paid-search benchmark, measured by DV Rockerbox, according to the October 5 post.
  • Dose (wellness): WorkMagic’s geo-lift study “estimated 2.3x more incremental orders” than last-click attribution captured, OpenAI said.

Newegg’s figure also appears unnamed in OpenAI’s August 31 post, as “an ecommerce advertiser achieving a 3x return on ad spend across its campaigns over 28 days.”

Can these numbers be compared?

Mostly not. Each story uses a different yardstick:

  • Attribution: Loop counts post-click orders and excludes view-through conversions. Talkpal and LegalNature use last-click. OpenAI’s own reporting lets advertisers add a one-day view-through window, and its October post says “52.7% of eligible one-day view-through conversions occurred within an hour of the matched ad impression.”
  • Benchmarks: Talkpal compares against one native channel, Jotform against paid search and paid social, Stream against channels it does not name, and WeightWatchers against a “blended” paid search figure.
  • Duration: Loop’s five feed campaigns launched between August 26 and September 8, so some had run for days, not weeks, when the window closed on September 8.
  • Selection: All seven were picked by OpenAI for its sales site. None reports a failed test.

There is also a gap in what is measured. Jotform says its CTR improved by 65% after changing creative, and that moving from reach to click campaigns cut CPC by “4-5x.” Neither story gives the CTR or CPC itself.

What do independent numbers show?

Three companies have posted their own figures on X. None has been independently verified:

  • Similarweb, May 3, 2026: “Overall CTR: 0.68%. Top quartile: 1%. Best brands: 1.57%. Peak: 5.4%.” It compared that with “around 0.35%” for display ads. The post does not give a sample size or date range, and it came two days before CPC bidding launched.
  • SE Ranking, September 17, 2026: 97,131 impressions, 1,263 clicks, a 1.30% CTR and a $3.16 CPC over about two weeks, across the US, Canada, Australia and New Zealand. That is roughly $3,990 in spend, by our calculation. The result: “Almost no sign-ups.”
  • Common Thread Collective, July 2, 2026: One account went from $7 to $1,005 a day over 34 days, with “~5x ROAS” and a $4.41 CPC. The agency said the product had an average order value above $1,000.

SE Ranking, a search tool company, gave a reason for its result in a longer write-up. “Ads reach Free and Go users only,” it wrote, while its buyers, SEO leads and agency owners, are “disproportionately likely to be on a paid, ad-free plan.” It also said it launched early, bid manual CPC and “had no in-platform conversion data to optimize against.”

Do ads change what ChatGPT cites?

Not in SE Ranking’s data. It recorded 50,006 commercial prompts across 20 niches in the US, of which 12,974 returned an ad, an ad rate of 25.94%. In those placements:

  • Domain citations: The advertiser’s domain appeared among the answer’s cited sources in 3.63% of cases.
  • Exact URL: The advertised page itself was cited in 0.09%, or 12 cases.
  • Mentions: The brand was named in the answer in 4.44%.
  • Relevance: 14.35% of ads were no closer to their prompt than a randomly paired ad, rising to 51.1% in relationships and 54.2% in news and politics, by SE Ranking’s measure.

SE Ranking also found ads on 28.69% of healthcare prompts. OpenAI’s testing post says ads are not eligible near health topics; SE Ranking’s study does not say what kind of healthcare prompts these were.

Background

  • January 16, 2026: OpenAI said it would test ads for US users on Free and Go.
  • February 9, 2026: The US test began, sold on a CPM basis.
  • March 26, 2026: OpenAI reported “no impact on consumer trust metrics” and “low dismissal rates,” without figures, and announced Canada, Australia and New Zealand.
  • May 5, 2026: CPC bidding, agency and technology partners, and self-serve Ads Manager in beta.
  • August 11 and 18, 2026: Launches in five more markets, then 31 European markets.
  • August 31, 2026: $1 billion run rate; self-serve opens in India, Europe, the Middle East and North Africa.
  • September 16, 2026: Sponsored Agents enter testing with select US advertisers; HubSpot and Shopify integrations launch.
  • October 5, 2026: A visual ad format for image generation, and new measurement partners.

What this means for AEO

In my view, the most useful number here isn’t OpenAI’s. It’s SE Ranking’s 3.63%. The ad slot and the answer above it are separate, and buying one does not buy the other. That fits what I found when I ran five shopping queries in ChatGPT: brands get named in answers through third-party sources, not their own pages.

The second point is audience. Ads only reach Free and Go users, so if your buyers pay for ChatGPT, they won’t see you there. For them, being cited in the answer is still the only way in. I’d test ads for ecommerce and consumer apps, where every published success sits. For B2B, check first whether your buyers are even in the eligible pool.

And treat the case studies as a sales deck. They are real numbers from real advertisers, but OpenAI picked them, and none uses the same yardstick. Until OpenAI publishes a CTR, the only absolute ones come from companies selling something of their own.

Questions OpenAI has not answered:

  • Benchmarks: What are the platform’s average CTR, CPC and conversion rate, by market?
  • Run rate: Which date did OpenAI count its “less than 200 days” from?
  • Trust: What were the trust and dismissal-rate figures behind its March update?
  • Health: How do ads appear on health-related prompts if health topics are excluded?
  • Reach: What share of ChatGPT’s 1.2 billion weekly users are on Free or Go and eligible to see ads?

Reference

Sandeep Amar, “Google’s AI Optimization Guide Is Not the AEO Bible. It’s a Press Release!”: https://sandeepamar.in/googles-ai-optimization-guide-is-not-the-aeo-bible-its-a-press-release/