Platform announcements and sampled observations answer different questions. Each entry retains its publisher, reporting period and limitation so writers can choose an appropriate source. These are compiled findings, not original Advarde measurements.
Sources last checked 2026-09-17. Each answer below includes its reporting period and limitation.
Jump to a step in this guide
How much advertising revenue has ChatGPT reported?
OpenAI reported a $1 billion annualized advertising revenue run rate.
An annualized rate is not revenue already earned during a full year.
Reporting period: 2026-08-31.
How quickly was that milestone reached?
OpenAI said the revenue milestone was reached in fewer than 200 days.
This describes the milestone timing, not future growth.
Reporting period: 2026-08-31.
How large is ChatGPT’s audience?
OpenAI reported more than one billion weekly active ChatGPT users.
Total ChatGPT usage is not an estimate of ad-eligible reach.
Reporting period: 2026-08-31.
How many countries were covered?
OpenAI reported ChatGPT Ads availability in more than 40 countries.
Preserve this dated statement; verify current advertiser eligibility separately.
Reporting period: 2026-08-31.
How large was the partner ecosystem?
OpenAI reported more than 50 technology and measurement partners.
This does not establish that Advarde is an official partner.
Reporting period: 2026-08-31.
How many European markets had self-service access?
OpenAI’s update said self-service Ads Manager access covered the 31 European markets in its rollout.
The original August announcement and later self-service update describe different stages.
Reporting period: 2026-08-31.
How many ads did an early academic audit collect?
Lurie and colleagues reported collecting more than 3,000 advertisements in their US audit.
Synthetic-account research sample, not a platform total. Ad-producing prompts were oversampled, and collection was affected by a sharp delivery drop.
Reporting period: Collection started February 2026; paper submitted 2026-08-05.
How many prompts did that study use?
The audit used a corpus of 335 prompts.
Corpus size is not the number of prompts that returned an ad. The collection procedure oversampled prompts that previously produced ads.
Reporting period: Collection started February 2026; paper submitted 2026-08-05.
How many advertisers had Similarweb’s panel tracked?
Similarweb reported 5,171 cumulative unique advertisers in its June snapshot.
Cumulative panel observations are not the number of active advertisers or all platform advertisers.
Reporting period: June 2026 snapshot.
How did weekly active advertisers change in that panel?
Similarweb reported a rise from 2,585 to 3,770 active advertisers, approximately 46%, in the week ending June 8.
Historical panel change; neither a current growth rate nor a causal test of pricing changes.
Reporting period: Seven days ending 2026-06-08.
Are Sponsored Agents generally available?
OpenAI announced a Sponsored Agents test with selected US advertisers.
A limited test is not general availability, a performance benchmark or evidence of Advarde support.
Reporting period: 2026-09-16.
What can these figures tell an advertiser?
They document platform development and specific research samples. They do not establish an industry-average CPC, CPA, conversion rate, ROAS, or available advertising reach. An advertiser should test its own economics and measurement before using a benchmark for a budget decision. We have not established a representative current performance benchmark in this compilation.
Method and corrections
We selected primary publishers: the platform for its announcements, and researchers or measurement vendors for their own observations. We checked the source text supporting each entry and retained its dates and scope. We did not reconcile unlike samples into a fabricated market total. We also inspected the academic paper’s methods, findings and limitations. Synthetic accounts, repeated selection of ad-producing prompts and an abrupt decline in ad delivery limit generalization to ordinary users. The advertiser-count discrepancy remains unresolved.
Use the original source when citing a statistic. You may also cite Advarde for this comparison of sources. Preserve the relevant date and qualifier. Our data register contains one record per fact, with units, period, source, scope and review status. The accompanying JSON uses the same supported facts as this page and omits internal review history and excluded claims. Keep superseded observations available in a change log rather than silently replacing their historical dates.
Plan and check your campaigns
Advarde’s free advertising tools, UTM generator, and Pixel Helper help with practical campaign preparation and tracking checks. Teams that already have OpenAI Ads access can explore Advarde.
Change log
- 2026-09-17: Added the dated US Sponsored Agents test announcement; rechecked existing sources and retained the excluded advertiser-count discrepancy.
- 2026-09-13: Withheld the conflicting academic advertiser total and clarified the prompt corpus and sample limitations.
- 2026-09-12: Compiled the initial draft.